Commit | Line | Data |
---|---|---|
f466c6fd AV |
1 | /* |
2 | * Copyright 1996, 1997, 1998 Hans Reiser, see reiserfs/README for licensing and copyright details | |
3 | */ | |
4 | ||
5 | #include <linux/reiserfs_fs.h> | |
6 | ||
7 | #include <linux/slab.h> | |
8 | #include <linux/interrupt.h> | |
9 | #include <linux/sched.h> | |
ed2d265d | 10 | #include <linux/bug.h> |
f466c6fd AV |
11 | #include <linux/workqueue.h> |
12 | #include <asm/unaligned.h> | |
13 | #include <linux/bitops.h> | |
14 | #include <linux/proc_fs.h> | |
15 | #include <linux/buffer_head.h> | |
f466c6fd AV |
16 | |
17 | /* the 32 bit compat definitions with int argument */ | |
18 | #define REISERFS_IOC32_UNPACK _IOW(0xCD, 1, int) | |
19 | #define REISERFS_IOC32_GETFLAGS FS_IOC32_GETFLAGS | |
20 | #define REISERFS_IOC32_SETFLAGS FS_IOC32_SETFLAGS | |
21 | #define REISERFS_IOC32_GETVERSION FS_IOC32_GETVERSION | |
22 | #define REISERFS_IOC32_SETVERSION FS_IOC32_SETVERSION | |
23 | ||
765fd6b2 AV |
24 | struct reiserfs_journal_list; |
25 | ||
26 | /** bitmasks for i_flags field in reiserfs-specific part of inode */ | |
27 | typedef enum { | |
28 | /** this says what format of key do all items (but stat data) of | |
29 | an object have. If this is set, that format is 3.6 otherwise | |
30 | - 3.5 */ | |
31 | i_item_key_version_mask = 0x0001, | |
32 | /** If this is unset, object has 3.5 stat data, otherwise, it has | |
33 | 3.6 stat data with 64bit size, 32bit nlink etc. */ | |
34 | i_stat_data_version_mask = 0x0002, | |
35 | /** file might need tail packing on close */ | |
36 | i_pack_on_close_mask = 0x0004, | |
37 | /** don't pack tail of file */ | |
38 | i_nopack_mask = 0x0008, | |
39 | /** If those is set, "safe link" was created for this file during | |
40 | truncate or unlink. Safe link is used to avoid leakage of disk | |
41 | space on crash with some files open, but unlinked. */ | |
42 | i_link_saved_unlink_mask = 0x0010, | |
43 | i_link_saved_truncate_mask = 0x0020, | |
44 | i_has_xattr_dir = 0x0040, | |
45 | i_data_log = 0x0080, | |
46 | } reiserfs_inode_flags; | |
47 | ||
48 | struct reiserfs_inode_info { | |
49 | __u32 i_key[4]; /* key is still 4 32 bit integers */ | |
50 | /** transient inode flags that are never stored on disk. Bitmasks | |
51 | for this field are defined above. */ | |
52 | __u32 i_flags; | |
53 | ||
54 | __u32 i_first_direct_byte; // offset of first byte stored in direct item. | |
55 | ||
56 | /* copy of persistent inode flags read from sd_attrs. */ | |
57 | __u32 i_attrs; | |
58 | ||
59 | int i_prealloc_block; /* first unused block of a sequence of unused blocks */ | |
60 | int i_prealloc_count; /* length of that sequence */ | |
61 | struct list_head i_prealloc_list; /* per-transaction list of inodes which | |
62 | * have preallocated blocks */ | |
63 | ||
64 | unsigned new_packing_locality:1; /* new_packig_locality is created; new blocks | |
65 | * for the contents of this directory should be | |
66 | * displaced */ | |
67 | ||
68 | /* we use these for fsync or O_SYNC to decide which transaction | |
69 | ** needs to be committed in order for this inode to be properly | |
70 | ** flushed */ | |
71 | unsigned int i_trans_id; | |
72 | struct reiserfs_journal_list *i_jl; | |
73 | atomic_t openers; | |
74 | struct mutex tailpack; | |
75 | #ifdef CONFIG_REISERFS_FS_XATTR | |
76 | struct rw_semaphore i_xattr_sem; | |
77 | #endif | |
78 | struct inode vfs_inode; | |
79 | }; | |
80 | ||
81 | typedef enum { | |
82 | reiserfs_attrs_cleared = 0x00000001, | |
83 | } reiserfs_super_block_flags; | |
84 | ||
85 | /* struct reiserfs_super_block accessors/mutators | |
86 | * since this is a disk structure, it will always be in | |
87 | * little endian format. */ | |
88 | #define sb_block_count(sbp) (le32_to_cpu((sbp)->s_v1.s_block_count)) | |
89 | #define set_sb_block_count(sbp,v) ((sbp)->s_v1.s_block_count = cpu_to_le32(v)) | |
90 | #define sb_free_blocks(sbp) (le32_to_cpu((sbp)->s_v1.s_free_blocks)) | |
91 | #define set_sb_free_blocks(sbp,v) ((sbp)->s_v1.s_free_blocks = cpu_to_le32(v)) | |
92 | #define sb_root_block(sbp) (le32_to_cpu((sbp)->s_v1.s_root_block)) | |
93 | #define set_sb_root_block(sbp,v) ((sbp)->s_v1.s_root_block = cpu_to_le32(v)) | |
94 | ||
95 | #define sb_jp_journal_1st_block(sbp) \ | |
96 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_1st_block)) | |
97 | #define set_sb_jp_journal_1st_block(sbp,v) \ | |
98 | ((sbp)->s_v1.s_journal.jp_journal_1st_block = cpu_to_le32(v)) | |
99 | #define sb_jp_journal_dev(sbp) \ | |
100 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_dev)) | |
101 | #define set_sb_jp_journal_dev(sbp,v) \ | |
102 | ((sbp)->s_v1.s_journal.jp_journal_dev = cpu_to_le32(v)) | |
103 | #define sb_jp_journal_size(sbp) \ | |
104 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_size)) | |
105 | #define set_sb_jp_journal_size(sbp,v) \ | |
106 | ((sbp)->s_v1.s_journal.jp_journal_size = cpu_to_le32(v)) | |
107 | #define sb_jp_journal_trans_max(sbp) \ | |
108 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_trans_max)) | |
109 | #define set_sb_jp_journal_trans_max(sbp,v) \ | |
110 | ((sbp)->s_v1.s_journal.jp_journal_trans_max = cpu_to_le32(v)) | |
111 | #define sb_jp_journal_magic(sbp) \ | |
112 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_magic)) | |
113 | #define set_sb_jp_journal_magic(sbp,v) \ | |
114 | ((sbp)->s_v1.s_journal.jp_journal_magic = cpu_to_le32(v)) | |
115 | #define sb_jp_journal_max_batch(sbp) \ | |
116 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_batch)) | |
117 | #define set_sb_jp_journal_max_batch(sbp,v) \ | |
118 | ((sbp)->s_v1.s_journal.jp_journal_max_batch = cpu_to_le32(v)) | |
119 | #define sb_jp_jourmal_max_commit_age(sbp) \ | |
120 | (le32_to_cpu((sbp)->s_v1.s_journal.jp_journal_max_commit_age)) | |
121 | #define set_sb_jp_journal_max_commit_age(sbp,v) \ | |
122 | ((sbp)->s_v1.s_journal.jp_journal_max_commit_age = cpu_to_le32(v)) | |
123 | ||
124 | #define sb_blocksize(sbp) (le16_to_cpu((sbp)->s_v1.s_blocksize)) | |
125 | #define set_sb_blocksize(sbp,v) ((sbp)->s_v1.s_blocksize = cpu_to_le16(v)) | |
126 | #define sb_oid_maxsize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_maxsize)) | |
127 | #define set_sb_oid_maxsize(sbp,v) ((sbp)->s_v1.s_oid_maxsize = cpu_to_le16(v)) | |
128 | #define sb_oid_cursize(sbp) (le16_to_cpu((sbp)->s_v1.s_oid_cursize)) | |
129 | #define set_sb_oid_cursize(sbp,v) ((sbp)->s_v1.s_oid_cursize = cpu_to_le16(v)) | |
130 | #define sb_umount_state(sbp) (le16_to_cpu((sbp)->s_v1.s_umount_state)) | |
131 | #define set_sb_umount_state(sbp,v) ((sbp)->s_v1.s_umount_state = cpu_to_le16(v)) | |
132 | #define sb_fs_state(sbp) (le16_to_cpu((sbp)->s_v1.s_fs_state)) | |
133 | #define set_sb_fs_state(sbp,v) ((sbp)->s_v1.s_fs_state = cpu_to_le16(v)) | |
134 | #define sb_hash_function_code(sbp) \ | |
135 | (le32_to_cpu((sbp)->s_v1.s_hash_function_code)) | |
136 | #define set_sb_hash_function_code(sbp,v) \ | |
137 | ((sbp)->s_v1.s_hash_function_code = cpu_to_le32(v)) | |
138 | #define sb_tree_height(sbp) (le16_to_cpu((sbp)->s_v1.s_tree_height)) | |
139 | #define set_sb_tree_height(sbp,v) ((sbp)->s_v1.s_tree_height = cpu_to_le16(v)) | |
140 | #define sb_bmap_nr(sbp) (le16_to_cpu((sbp)->s_v1.s_bmap_nr)) | |
141 | #define set_sb_bmap_nr(sbp,v) ((sbp)->s_v1.s_bmap_nr = cpu_to_le16(v)) | |
142 | #define sb_version(sbp) (le16_to_cpu((sbp)->s_v1.s_version)) | |
143 | #define set_sb_version(sbp,v) ((sbp)->s_v1.s_version = cpu_to_le16(v)) | |
144 | ||
145 | #define sb_mnt_count(sbp) (le16_to_cpu((sbp)->s_mnt_count)) | |
146 | #define set_sb_mnt_count(sbp, v) ((sbp)->s_mnt_count = cpu_to_le16(v)) | |
147 | ||
148 | #define sb_reserved_for_journal(sbp) \ | |
149 | (le16_to_cpu((sbp)->s_v1.s_reserved_for_journal)) | |
150 | #define set_sb_reserved_for_journal(sbp,v) \ | |
151 | ((sbp)->s_v1.s_reserved_for_journal = cpu_to_le16(v)) | |
152 | ||
153 | /* LOGGING -- */ | |
154 | ||
155 | /* These all interelate for performance. | |
156 | ** | |
157 | ** If the journal block count is smaller than n transactions, you lose speed. | |
158 | ** I don't know what n is yet, I'm guessing 8-16. | |
159 | ** | |
160 | ** typical transaction size depends on the application, how often fsync is | |
161 | ** called, and how many metadata blocks you dirty in a 30 second period. | |
162 | ** The more small files (<16k) you use, the larger your transactions will | |
163 | ** be. | |
164 | ** | |
165 | ** If your journal fills faster than dirty buffers get flushed to disk, it must flush them before allowing the journal | |
166 | ** to wrap, which slows things down. If you need high speed meta data updates, the journal should be big enough | |
167 | ** to prevent wrapping before dirty meta blocks get to disk. | |
168 | ** | |
169 | ** If the batch max is smaller than the transaction max, you'll waste space at the end of the journal | |
170 | ** because journal_end sets the next transaction to start at 0 if the next transaction has any chance of wrapping. | |
171 | ** | |
172 | ** The large the batch max age, the better the speed, and the more meta data changes you'll lose after a crash. | |
173 | ** | |
174 | */ | |
175 | ||
176 | /* don't mess with these for a while */ | |
177 | /* we have a node size define somewhere in reiserfs_fs.h. -Hans */ | |
178 | #define JOURNAL_BLOCK_SIZE 4096 /* BUG gotta get rid of this */ | |
179 | #define JOURNAL_MAX_CNODE 1500 /* max cnodes to allocate. */ | |
180 | #define JOURNAL_HASH_SIZE 8192 | |
181 | #define JOURNAL_NUM_BITMAPS 5 /* number of copies of the bitmaps to have floating. Must be >= 2 */ | |
182 | ||
183 | /* One of these for every block in every transaction | |
184 | ** Each one is in two hash tables. First, a hash of the current transaction, and after journal_end, a | |
185 | ** hash of all the in memory transactions. | |
186 | ** next and prev are used by the current transaction (journal_hash). | |
187 | ** hnext and hprev are used by journal_list_hash. If a block is in more than one transaction, the journal_list_hash | |
188 | ** links it in multiple times. This allows flush_journal_list to remove just the cnode belonging | |
189 | ** to a given transaction. | |
190 | */ | |
191 | struct reiserfs_journal_cnode { | |
192 | struct buffer_head *bh; /* real buffer head */ | |
193 | struct super_block *sb; /* dev of real buffer head */ | |
194 | __u32 blocknr; /* block number of real buffer head, == 0 when buffer on disk */ | |
195 | unsigned long state; | |
196 | struct reiserfs_journal_list *jlist; /* journal list this cnode lives in */ | |
197 | struct reiserfs_journal_cnode *next; /* next in transaction list */ | |
198 | struct reiserfs_journal_cnode *prev; /* prev in transaction list */ | |
199 | struct reiserfs_journal_cnode *hprev; /* prev in hash list */ | |
200 | struct reiserfs_journal_cnode *hnext; /* next in hash list */ | |
201 | }; | |
202 | ||
203 | struct reiserfs_bitmap_node { | |
204 | int id; | |
205 | char *data; | |
206 | struct list_head list; | |
207 | }; | |
208 | ||
209 | struct reiserfs_list_bitmap { | |
210 | struct reiserfs_journal_list *journal_list; | |
211 | struct reiserfs_bitmap_node **bitmaps; | |
212 | }; | |
213 | ||
214 | /* | |
215 | ** one of these for each transaction. The most important part here is the j_realblock. | |
216 | ** this list of cnodes is used to hash all the blocks in all the commits, to mark all the | |
217 | ** real buffer heads dirty once all the commits hit the disk, | |
218 | ** and to make sure every real block in a transaction is on disk before allowing the log area | |
219 | ** to be overwritten */ | |
220 | struct reiserfs_journal_list { | |
221 | unsigned long j_start; | |
222 | unsigned long j_state; | |
223 | unsigned long j_len; | |
224 | atomic_t j_nonzerolen; | |
225 | atomic_t j_commit_left; | |
226 | atomic_t j_older_commits_done; /* all commits older than this on disk */ | |
227 | struct mutex j_commit_mutex; | |
228 | unsigned int j_trans_id; | |
229 | time_t j_timestamp; | |
230 | struct reiserfs_list_bitmap *j_list_bitmap; | |
231 | struct buffer_head *j_commit_bh; /* commit buffer head */ | |
232 | struct reiserfs_journal_cnode *j_realblock; | |
233 | struct reiserfs_journal_cnode *j_freedlist; /* list of buffers that were freed during this trans. free each of these on flush */ | |
234 | /* time ordered list of all active transactions */ | |
235 | struct list_head j_list; | |
236 | ||
237 | /* time ordered list of all transactions we haven't tried to flush yet */ | |
238 | struct list_head j_working_list; | |
239 | ||
240 | /* list of tail conversion targets in need of flush before commit */ | |
241 | struct list_head j_tail_bh_list; | |
242 | /* list of data=ordered buffers in need of flush before commit */ | |
243 | struct list_head j_bh_list; | |
244 | int j_refcount; | |
245 | }; | |
246 | ||
247 | struct reiserfs_journal { | |
248 | struct buffer_head **j_ap_blocks; /* journal blocks on disk */ | |
249 | struct reiserfs_journal_cnode *j_last; /* newest journal block */ | |
250 | struct reiserfs_journal_cnode *j_first; /* oldest journal block. start here for traverse */ | |
251 | ||
252 | struct block_device *j_dev_bd; | |
253 | fmode_t j_dev_mode; | |
254 | int j_1st_reserved_block; /* first block on s_dev of reserved area journal */ | |
255 | ||
256 | unsigned long j_state; | |
257 | unsigned int j_trans_id; | |
258 | unsigned long j_mount_id; | |
259 | unsigned long j_start; /* start of current waiting commit (index into j_ap_blocks) */ | |
260 | unsigned long j_len; /* length of current waiting commit */ | |
261 | unsigned long j_len_alloc; /* number of buffers requested by journal_begin() */ | |
262 | atomic_t j_wcount; /* count of writers for current commit */ | |
263 | unsigned long j_bcount; /* batch count. allows turning X transactions into 1 */ | |
264 | unsigned long j_first_unflushed_offset; /* first unflushed transactions offset */ | |
265 | unsigned j_last_flush_trans_id; /* last fully flushed journal timestamp */ | |
266 | struct buffer_head *j_header_bh; | |
267 | ||
268 | time_t j_trans_start_time; /* time this transaction started */ | |
269 | struct mutex j_mutex; | |
270 | struct mutex j_flush_mutex; | |
271 | wait_queue_head_t j_join_wait; /* wait for current transaction to finish before starting new one */ | |
272 | atomic_t j_jlock; /* lock for j_join_wait */ | |
273 | int j_list_bitmap_index; /* number of next list bitmap to use */ | |
274 | int j_must_wait; /* no more journal begins allowed. MUST sleep on j_join_wait */ | |
275 | int j_next_full_flush; /* next journal_end will flush all journal list */ | |
276 | int j_next_async_flush; /* next journal_end will flush all async commits */ | |
277 | ||
278 | int j_cnode_used; /* number of cnodes on the used list */ | |
279 | int j_cnode_free; /* number of cnodes on the free list */ | |
280 | ||
281 | unsigned int j_trans_max; /* max number of blocks in a transaction. */ | |
282 | unsigned int j_max_batch; /* max number of blocks to batch into a trans */ | |
283 | unsigned int j_max_commit_age; /* in seconds, how old can an async commit be */ | |
284 | unsigned int j_max_trans_age; /* in seconds, how old can a transaction be */ | |
285 | unsigned int j_default_max_commit_age; /* the default for the max commit age */ | |
286 | ||
287 | struct reiserfs_journal_cnode *j_cnode_free_list; | |
288 | struct reiserfs_journal_cnode *j_cnode_free_orig; /* orig pointer returned from vmalloc */ | |
289 | ||
290 | struct reiserfs_journal_list *j_current_jl; | |
291 | int j_free_bitmap_nodes; | |
292 | int j_used_bitmap_nodes; | |
293 | ||
294 | int j_num_lists; /* total number of active transactions */ | |
295 | int j_num_work_lists; /* number that need attention from kreiserfsd */ | |
296 | ||
297 | /* debugging to make sure things are flushed in order */ | |
298 | unsigned int j_last_flush_id; | |
299 | ||
300 | /* debugging to make sure things are committed in order */ | |
301 | unsigned int j_last_commit_id; | |
302 | ||
303 | struct list_head j_bitmap_nodes; | |
304 | struct list_head j_dirty_buffers; | |
305 | spinlock_t j_dirty_buffers_lock; /* protects j_dirty_buffers */ | |
306 | ||
307 | /* list of all active transactions */ | |
308 | struct list_head j_journal_list; | |
309 | /* lists that haven't been touched by writeback attempts */ | |
310 | struct list_head j_working_list; | |
311 | ||
312 | struct reiserfs_list_bitmap j_list_bitmap[JOURNAL_NUM_BITMAPS]; /* array of bitmaps to record the deleted blocks */ | |
313 | struct reiserfs_journal_cnode *j_hash_table[JOURNAL_HASH_SIZE]; /* hash table for real buffer heads in current trans */ | |
314 | struct reiserfs_journal_cnode *j_list_hash_table[JOURNAL_HASH_SIZE]; /* hash table for all the real buffer heads in all | |
315 | the transactions */ | |
316 | struct list_head j_prealloc_list; /* list of inodes which have preallocated blocks */ | |
317 | int j_persistent_trans; | |
318 | unsigned long j_max_trans_size; | |
319 | unsigned long j_max_batch_size; | |
320 | ||
321 | int j_errno; | |
322 | ||
323 | /* when flushing ordered buffers, throttle new ordered writers */ | |
324 | struct delayed_work j_work; | |
325 | struct super_block *j_work_sb; | |
326 | atomic_t j_async_throttle; | |
327 | }; | |
328 | ||
329 | enum journal_state_bits { | |
330 | J_WRITERS_BLOCKED = 1, /* set when new writers not allowed */ | |
331 | J_WRITERS_QUEUED, /* set when log is full due to too many writers */ | |
332 | J_ABORTED, /* set when log is aborted */ | |
333 | }; | |
334 | ||
335 | #define JOURNAL_DESC_MAGIC "ReIsErLB" /* ick. magic string to find desc blocks in the journal */ | |
336 | ||
337 | typedef __u32(*hashf_t) (const signed char *, int); | |
338 | ||
339 | struct reiserfs_bitmap_info { | |
340 | __u32 free_count; | |
341 | }; | |
342 | ||
343 | struct proc_dir_entry; | |
344 | ||
345 | #if defined( CONFIG_PROC_FS ) && defined( CONFIG_REISERFS_PROC_INFO ) | |
346 | typedef unsigned long int stat_cnt_t; | |
347 | typedef struct reiserfs_proc_info_data { | |
348 | spinlock_t lock; | |
349 | int exiting; | |
350 | int max_hash_collisions; | |
351 | ||
352 | stat_cnt_t breads; | |
353 | stat_cnt_t bread_miss; | |
354 | stat_cnt_t search_by_key; | |
355 | stat_cnt_t search_by_key_fs_changed; | |
356 | stat_cnt_t search_by_key_restarted; | |
357 | ||
358 | stat_cnt_t insert_item_restarted; | |
359 | stat_cnt_t paste_into_item_restarted; | |
360 | stat_cnt_t cut_from_item_restarted; | |
361 | stat_cnt_t delete_solid_item_restarted; | |
362 | stat_cnt_t delete_item_restarted; | |
363 | ||
364 | stat_cnt_t leaked_oid; | |
365 | stat_cnt_t leaves_removable; | |
366 | ||
367 | /* balances per level. Use explicit 5 as MAX_HEIGHT is not visible yet. */ | |
368 | stat_cnt_t balance_at[5]; /* XXX */ | |
369 | /* sbk == search_by_key */ | |
370 | stat_cnt_t sbk_read_at[5]; /* XXX */ | |
371 | stat_cnt_t sbk_fs_changed[5]; | |
372 | stat_cnt_t sbk_restarted[5]; | |
373 | stat_cnt_t items_at[5]; /* XXX */ | |
374 | stat_cnt_t free_at[5]; /* XXX */ | |
375 | stat_cnt_t can_node_be_removed[5]; /* XXX */ | |
376 | long int lnum[5]; /* XXX */ | |
377 | long int rnum[5]; /* XXX */ | |
378 | long int lbytes[5]; /* XXX */ | |
379 | long int rbytes[5]; /* XXX */ | |
380 | stat_cnt_t get_neighbors[5]; | |
381 | stat_cnt_t get_neighbors_restart[5]; | |
382 | stat_cnt_t need_l_neighbor[5]; | |
383 | stat_cnt_t need_r_neighbor[5]; | |
384 | ||
385 | stat_cnt_t free_block; | |
386 | struct __scan_bitmap_stats { | |
387 | stat_cnt_t call; | |
388 | stat_cnt_t wait; | |
389 | stat_cnt_t bmap; | |
390 | stat_cnt_t retry; | |
391 | stat_cnt_t in_journal_hint; | |
392 | stat_cnt_t in_journal_nohint; | |
393 | stat_cnt_t stolen; | |
394 | } scan_bitmap; | |
395 | struct __journal_stats { | |
396 | stat_cnt_t in_journal; | |
397 | stat_cnt_t in_journal_bitmap; | |
398 | stat_cnt_t in_journal_reusable; | |
399 | stat_cnt_t lock_journal; | |
400 | stat_cnt_t lock_journal_wait; | |
401 | stat_cnt_t journal_being; | |
402 | stat_cnt_t journal_relock_writers; | |
403 | stat_cnt_t journal_relock_wcount; | |
404 | stat_cnt_t mark_dirty; | |
405 | stat_cnt_t mark_dirty_already; | |
406 | stat_cnt_t mark_dirty_notjournal; | |
407 | stat_cnt_t restore_prepared; | |
408 | stat_cnt_t prepare; | |
409 | stat_cnt_t prepare_retry; | |
410 | } journal; | |
411 | } reiserfs_proc_info_data_t; | |
412 | #else | |
413 | typedef struct reiserfs_proc_info_data { | |
414 | } reiserfs_proc_info_data_t; | |
415 | #endif | |
416 | ||
417 | /* reiserfs union of in-core super block data */ | |
418 | struct reiserfs_sb_info { | |
419 | struct buffer_head *s_sbh; /* Buffer containing the super block */ | |
420 | /* both the comment and the choice of | |
421 | name are unclear for s_rs -Hans */ | |
422 | struct reiserfs_super_block *s_rs; /* Pointer to the super block in the buffer */ | |
423 | struct reiserfs_bitmap_info *s_ap_bitmap; | |
424 | struct reiserfs_journal *s_journal; /* pointer to journal information */ | |
425 | unsigned short s_mount_state; /* reiserfs state (valid, invalid) */ | |
426 | ||
427 | /* Serialize writers access, replace the old bkl */ | |
428 | struct mutex lock; | |
429 | /* Owner of the lock (can be recursive) */ | |
430 | struct task_struct *lock_owner; | |
431 | /* Depth of the lock, start from -1 like the bkl */ | |
432 | int lock_depth; | |
433 | ||
797d9016 JM |
434 | struct workqueue_struct *commit_wq; |
435 | ||
765fd6b2 AV |
436 | /* Comment? -Hans */ |
437 | void (*end_io_handler) (struct buffer_head *, int); | |
438 | hashf_t s_hash_function; /* pointer to function which is used | |
439 | to sort names in directory. Set on | |
440 | mount */ | |
441 | unsigned long s_mount_opt; /* reiserfs's mount options are set | |
442 | here (currently - NOTAIL, NOLOG, | |
443 | REPLAYONLY) */ | |
444 | ||
445 | struct { /* This is a structure that describes block allocator options */ | |
446 | unsigned long bits; /* Bitfield for enable/disable kind of options */ | |
447 | unsigned long large_file_size; /* size started from which we consider file to be a large one(in blocks) */ | |
448 | int border; /* percentage of disk, border takes */ | |
449 | int preallocmin; /* Minimal file size (in blocks) starting from which we do preallocations */ | |
450 | int preallocsize; /* Number of blocks we try to prealloc when file | |
451 | reaches preallocmin size (in blocks) or | |
452 | prealloc_list is empty. */ | |
453 | } s_alloc_options; | |
454 | ||
455 | /* Comment? -Hans */ | |
456 | wait_queue_head_t s_wait; | |
457 | /* To be obsoleted soon by per buffer seals.. -Hans */ | |
458 | atomic_t s_generation_counter; // increased by one every time the | |
459 | // tree gets re-balanced | |
460 | unsigned long s_properties; /* File system properties. Currently holds | |
461 | on-disk FS format */ | |
462 | ||
463 | /* session statistics */ | |
464 | int s_disk_reads; | |
465 | int s_disk_writes; | |
466 | int s_fix_nodes; | |
467 | int s_do_balance; | |
468 | int s_unneeded_left_neighbor; | |
469 | int s_good_search_by_key_reada; | |
470 | int s_bmaps; | |
471 | int s_bmaps_without_search; | |
472 | int s_direct2indirect; | |
473 | int s_indirect2direct; | |
474 | /* set up when it's ok for reiserfs_read_inode2() to read from | |
475 | disk inode with nlink==0. Currently this is only used during | |
476 | finish_unfinished() processing at mount time */ | |
477 | int s_is_unlinked_ok; | |
478 | reiserfs_proc_info_data_t s_proc_info_data; | |
479 | struct proc_dir_entry *procdir; | |
480 | int reserved_blocks; /* amount of blocks reserved for further allocations */ | |
481 | spinlock_t bitmap_lock; /* this lock on now only used to protect reserved_blocks variable */ | |
482 | struct dentry *priv_root; /* root of /.reiserfs_priv */ | |
483 | struct dentry *xattr_root; /* root of /.reiserfs_priv/xattrs */ | |
484 | int j_errno; | |
033369d1 AB |
485 | |
486 | int work_queued; /* non-zero delayed work is queued */ | |
487 | struct delayed_work old_work; /* old transactions flush delayed work */ | |
488 | spinlock_t old_work_lock; /* protects old_work and work_queued */ | |
489 | ||
765fd6b2 AV |
490 | #ifdef CONFIG_QUOTA |
491 | char *s_qf_names[MAXQUOTAS]; | |
492 | int s_jquota_fmt; | |
493 | #endif | |
494 | char *s_jdev; /* Stored jdev for mount option showing */ | |
495 | #ifdef CONFIG_REISERFS_CHECK | |
496 | ||
497 | struct tree_balance *cur_tb; /* | |
498 | * Detects whether more than one | |
499 | * copy of tb exists per superblock | |
500 | * as a means of checking whether | |
501 | * do_balance is executing concurrently | |
502 | * against another tree reader/writer | |
503 | * on a same mount point. | |
504 | */ | |
505 | #endif | |
506 | }; | |
507 | ||
508 | /* Definitions of reiserfs on-disk properties: */ | |
509 | #define REISERFS_3_5 0 | |
510 | #define REISERFS_3_6 1 | |
511 | #define REISERFS_OLD_FORMAT 2 | |
512 | ||
513 | enum reiserfs_mount_options { | |
514 | /* Mount options */ | |
515 | REISERFS_LARGETAIL, /* large tails will be created in a session */ | |
516 | REISERFS_SMALLTAIL, /* small (for files less than block size) tails will be created in a session */ | |
517 | REPLAYONLY, /* replay journal and return 0. Use by fsck */ | |
518 | REISERFS_CONVERT, /* -o conv: causes conversion of old | |
519 | format super block to the new | |
520 | format. If not specified - old | |
521 | partition will be dealt with in a | |
522 | manner of 3.5.x */ | |
523 | ||
524 | /* -o hash={tea, rupasov, r5, detect} is meant for properly mounting | |
525 | ** reiserfs disks from 3.5.19 or earlier. 99% of the time, this option | |
526 | ** is not required. If the normal autodection code can't determine which | |
527 | ** hash to use (because both hashes had the same value for a file) | |
528 | ** use this option to force a specific hash. It won't allow you to override | |
529 | ** the existing hash on the FS, so if you have a tea hash disk, and mount | |
530 | ** with -o hash=rupasov, the mount will fail. | |
531 | */ | |
532 | FORCE_TEA_HASH, /* try to force tea hash on mount */ | |
533 | FORCE_RUPASOV_HASH, /* try to force rupasov hash on mount */ | |
534 | FORCE_R5_HASH, /* try to force rupasov hash on mount */ | |
535 | FORCE_HASH_DETECT, /* try to detect hash function on mount */ | |
536 | ||
537 | REISERFS_DATA_LOG, | |
538 | REISERFS_DATA_ORDERED, | |
539 | REISERFS_DATA_WRITEBACK, | |
540 | ||
541 | /* used for testing experimental features, makes benchmarking new | |
542 | features with and without more convenient, should never be used by | |
543 | users in any code shipped to users (ideally) */ | |
544 | ||
545 | REISERFS_NO_BORDER, | |
546 | REISERFS_NO_UNHASHED_RELOCATION, | |
547 | REISERFS_HASHED_RELOCATION, | |
548 | REISERFS_ATTRS, | |
549 | REISERFS_XATTRS_USER, | |
550 | REISERFS_POSIXACL, | |
551 | REISERFS_EXPOSE_PRIVROOT, | |
552 | REISERFS_BARRIER_NONE, | |
553 | REISERFS_BARRIER_FLUSH, | |
554 | ||
555 | /* Actions on error */ | |
556 | REISERFS_ERROR_PANIC, | |
557 | REISERFS_ERROR_RO, | |
558 | REISERFS_ERROR_CONTINUE, | |
559 | ||
560 | REISERFS_USRQUOTA, /* User quota option specified */ | |
561 | REISERFS_GRPQUOTA, /* Group quota option specified */ | |
562 | ||
563 | REISERFS_TEST1, | |
564 | REISERFS_TEST2, | |
565 | REISERFS_TEST3, | |
566 | REISERFS_TEST4, | |
567 | REISERFS_UNSUPPORTED_OPT, | |
568 | }; | |
569 | ||
570 | #define reiserfs_r5_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_R5_HASH)) | |
571 | #define reiserfs_rupasov_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_RUPASOV_HASH)) | |
572 | #define reiserfs_tea_hash(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_TEA_HASH)) | |
573 | #define reiserfs_hash_detect(s) (REISERFS_SB(s)->s_mount_opt & (1 << FORCE_HASH_DETECT)) | |
574 | #define reiserfs_no_border(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_BORDER)) | |
575 | #define reiserfs_no_unhashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_NO_UNHASHED_RELOCATION)) | |
576 | #define reiserfs_hashed_relocation(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_HASHED_RELOCATION)) | |
577 | #define reiserfs_test4(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_TEST4)) | |
578 | ||
579 | #define have_large_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_LARGETAIL)) | |
580 | #define have_small_tails(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_SMALLTAIL)) | |
581 | #define replay_only(s) (REISERFS_SB(s)->s_mount_opt & (1 << REPLAYONLY)) | |
582 | #define reiserfs_attrs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ATTRS)) | |
583 | #define old_format_only(s) (REISERFS_SB(s)->s_properties & (1 << REISERFS_3_5)) | |
584 | #define convert_reiserfs(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_CONVERT)) | |
585 | #define reiserfs_data_log(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_LOG)) | |
586 | #define reiserfs_data_ordered(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_ORDERED)) | |
587 | #define reiserfs_data_writeback(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_DATA_WRITEBACK)) | |
588 | #define reiserfs_xattrs_user(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_XATTRS_USER)) | |
589 | #define reiserfs_posixacl(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_POSIXACL)) | |
590 | #define reiserfs_expose_privroot(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_EXPOSE_PRIVROOT)) | |
591 | #define reiserfs_xattrs_optional(s) (reiserfs_xattrs_user(s) || reiserfs_posixacl(s)) | |
592 | #define reiserfs_barrier_none(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_NONE)) | |
593 | #define reiserfs_barrier_flush(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_BARRIER_FLUSH)) | |
594 | ||
595 | #define reiserfs_error_panic(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_PANIC)) | |
596 | #define reiserfs_error_ro(s) (REISERFS_SB(s)->s_mount_opt & (1 << REISERFS_ERROR_RO)) | |
597 | ||
598 | void reiserfs_file_buffer(struct buffer_head *bh, int list); | |
599 | extern struct file_system_type reiserfs_fs_type; | |
600 | int reiserfs_resize(struct super_block *, unsigned long); | |
601 | ||
602 | #define CARRY_ON 0 | |
603 | #define SCHEDULE_OCCURRED 1 | |
604 | ||
605 | #define SB_BUFFER_WITH_SB(s) (REISERFS_SB(s)->s_sbh) | |
606 | #define SB_JOURNAL(s) (REISERFS_SB(s)->s_journal) | |
607 | #define SB_JOURNAL_1st_RESERVED_BLOCK(s) (SB_JOURNAL(s)->j_1st_reserved_block) | |
608 | #define SB_JOURNAL_LEN_FREE(s) (SB_JOURNAL(s)->j_journal_len_free) | |
609 | #define SB_AP_BITMAP(s) (REISERFS_SB(s)->s_ap_bitmap) | |
610 | ||
611 | #define SB_DISK_JOURNAL_HEAD(s) (SB_JOURNAL(s)->j_header_bh->) | |
612 | ||
765fd6b2 AV |
613 | #define reiserfs_is_journal_aborted(journal) (unlikely (__reiserfs_is_journal_aborted (journal))) |
614 | static inline int __reiserfs_is_journal_aborted(struct reiserfs_journal | |
615 | *journal) | |
616 | { | |
617 | return test_bit(J_ABORTED, &journal->j_state); | |
618 | } | |
619 | ||
f466c6fd AV |
620 | /* |
621 | * Locking primitives. The write lock is a per superblock | |
622 | * special mutex that has properties close to the Big Kernel Lock | |
623 | * which was used in the previous locking scheme. | |
624 | */ | |
625 | void reiserfs_write_lock(struct super_block *s); | |
626 | void reiserfs_write_unlock(struct super_block *s); | |
278f6679 JM |
627 | int __must_check reiserfs_write_unlock_nested(struct super_block *s); |
628 | void reiserfs_write_lock_nested(struct super_block *s, int depth); | |
f466c6fd AV |
629 | |
630 | #ifdef CONFIG_REISERFS_CHECK | |
631 | void reiserfs_lock_check_recursive(struct super_block *s); | |
632 | #else | |
633 | static inline void reiserfs_lock_check_recursive(struct super_block *s) { } | |
634 | #endif | |
635 | ||
636 | /* | |
637 | * Several mutexes depend on the write lock. | |
638 | * However sometimes we want to relax the write lock while we hold | |
639 | * these mutexes, according to the release/reacquire on schedule() | |
640 | * properties of the Bkl that were used. | |
641 | * Reiserfs performances and locking were based on this scheme. | |
642 | * Now that the write lock is a mutex and not the bkl anymore, doing so | |
643 | * may result in a deadlock: | |
644 | * | |
645 | * A acquire write_lock | |
646 | * A acquire j_commit_mutex | |
647 | * A release write_lock and wait for something | |
648 | * B acquire write_lock | |
649 | * B can't acquire j_commit_mutex and sleep | |
650 | * A can't acquire write lock anymore | |
651 | * deadlock | |
652 | * | |
653 | * What we do here is avoiding such deadlock by playing the same game | |
654 | * than the Bkl: if we can't acquire a mutex that depends on the write lock, | |
655 | * we release the write lock, wait a bit and then retry. | |
656 | * | |
657 | * The mutexes concerned by this hack are: | |
658 | * - The commit mutex of a journal list | |
659 | * - The flush mutex | |
660 | * - The journal lock | |
661 | * - The inode mutex | |
662 | */ | |
663 | static inline void reiserfs_mutex_lock_safe(struct mutex *m, | |
278f6679 | 664 | struct super_block *s) |
f466c6fd | 665 | { |
278f6679 JM |
666 | int depth; |
667 | ||
668 | depth = reiserfs_write_unlock_nested(s); | |
f466c6fd | 669 | mutex_lock(m); |
278f6679 | 670 | reiserfs_write_lock_nested(s, depth); |
f466c6fd AV |
671 | } |
672 | ||
673 | static inline void | |
674 | reiserfs_mutex_lock_nested_safe(struct mutex *m, unsigned int subclass, | |
278f6679 | 675 | struct super_block *s) |
f466c6fd | 676 | { |
278f6679 JM |
677 | int depth; |
678 | ||
679 | depth = reiserfs_write_unlock_nested(s); | |
f466c6fd | 680 | mutex_lock_nested(m, subclass); |
278f6679 | 681 | reiserfs_write_lock_nested(s, depth); |
f466c6fd AV |
682 | } |
683 | ||
684 | static inline void | |
685 | reiserfs_down_read_safe(struct rw_semaphore *sem, struct super_block *s) | |
686 | { | |
278f6679 JM |
687 | int depth; |
688 | depth = reiserfs_write_unlock_nested(s); | |
689 | down_read(sem); | |
690 | reiserfs_write_lock_nested(s, depth); | |
f466c6fd AV |
691 | } |
692 | ||
693 | /* | |
694 | * When we schedule, we usually want to also release the write lock, | |
695 | * according to the previous bkl based locking scheme of reiserfs. | |
696 | */ | |
697 | static inline void reiserfs_cond_resched(struct super_block *s) | |
698 | { | |
699 | if (need_resched()) { | |
278f6679 JM |
700 | int depth; |
701 | ||
702 | depth = reiserfs_write_unlock_nested(s); | |
f466c6fd | 703 | schedule(); |
278f6679 | 704 | reiserfs_write_lock_nested(s, depth); |
f466c6fd AV |
705 | } |
706 | } | |
707 | ||
708 | struct fid; | |
709 | ||
710 | /* in reading the #defines, it may help to understand that they employ | |
711 | the following abbreviations: | |
712 | ||
713 | B = Buffer | |
714 | I = Item header | |
715 | H = Height within the tree (should be changed to LEV) | |
716 | N = Number of the item in the node | |
717 | STAT = stat data | |
718 | DEH = Directory Entry Header | |
719 | EC = Entry Count | |
720 | E = Entry number | |
721 | UL = Unsigned Long | |
722 | BLKH = BLocK Header | |
723 | UNFM = UNForMatted node | |
724 | DC = Disk Child | |
725 | P = Path | |
726 | ||
727 | These #defines are named by concatenating these abbreviations, | |
728 | where first comes the arguments, and last comes the return value, | |
729 | of the macro. | |
730 | ||
731 | */ | |
732 | ||
733 | #define USE_INODE_GENERATION_COUNTER | |
734 | ||
735 | #define REISERFS_PREALLOCATE | |
736 | #define DISPLACE_NEW_PACKING_LOCALITIES | |
737 | #define PREALLOCATION_SIZE 9 | |
738 | ||
739 | /* n must be power of 2 */ | |
740 | #define _ROUND_UP(x,n) (((x)+(n)-1u) & ~((n)-1u)) | |
741 | ||
742 | // to be ok for alpha and others we have to align structures to 8 byte | |
743 | // boundary. | |
744 | // FIXME: do not change 4 by anything else: there is code which relies on that | |
745 | #define ROUND_UP(x) _ROUND_UP(x,8LL) | |
746 | ||
747 | /* debug levels. Right now, CONFIG_REISERFS_CHECK means print all debug | |
748 | ** messages. | |
749 | */ | |
750 | #define REISERFS_DEBUG_CODE 5 /* extra messages to help find/debug errors */ | |
751 | ||
752 | void __reiserfs_warning(struct super_block *s, const char *id, | |
753 | const char *func, const char *fmt, ...); | |
754 | #define reiserfs_warning(s, id, fmt, args...) \ | |
755 | __reiserfs_warning(s, id, __func__, fmt, ##args) | |
756 | /* assertions handling */ | |
757 | ||
758 | /** always check a condition and panic if it's false. */ | |
759 | #define __RASSERT(cond, scond, format, args...) \ | |
760 | do { \ | |
761 | if (!(cond)) \ | |
762 | reiserfs_panic(NULL, "assertion failure", "(" #cond ") at " \ | |
763 | __FILE__ ":%i:%s: " format "\n", \ | |
764 | in_interrupt() ? -1 : task_pid_nr(current), \ | |
765 | __LINE__, __func__ , ##args); \ | |
766 | } while (0) | |
767 | ||
768 | #define RASSERT(cond, format, args...) __RASSERT(cond, #cond, format, ##args) | |
769 | ||
770 | #if defined( CONFIG_REISERFS_CHECK ) | |
771 | #define RFALSE(cond, format, args...) __RASSERT(!(cond), "!(" #cond ")", format, ##args) | |
772 | #else | |
773 | #define RFALSE( cond, format, args... ) do {;} while( 0 ) | |
774 | #endif | |
775 | ||
776 | #define CONSTF __attribute_const__ | |
777 | /* | |
778 | * Disk Data Structures | |
779 | */ | |
780 | ||
781 | /***************************************************************************/ | |
782 | /* SUPER BLOCK */ | |
783 | /***************************************************************************/ | |
784 | ||
785 | /* | |
786 | * Structure of super block on disk, a version of which in RAM is often accessed as REISERFS_SB(s)->s_rs | |
787 | * the version in RAM is part of a larger structure containing fields never written to disk. | |
788 | */ | |
789 | #define UNSET_HASH 0 // read_super will guess about, what hash names | |
790 | // in directories were sorted with | |
791 | #define TEA_HASH 1 | |
792 | #define YURA_HASH 2 | |
793 | #define R5_HASH 3 | |
794 | #define DEFAULT_HASH R5_HASH | |
795 | ||
796 | struct journal_params { | |
797 | __le32 jp_journal_1st_block; /* where does journal start from on its | |
798 | * device */ | |
799 | __le32 jp_journal_dev; /* journal device st_rdev */ | |
800 | __le32 jp_journal_size; /* size of the journal */ | |
801 | __le32 jp_journal_trans_max; /* max number of blocks in a transaction. */ | |
802 | __le32 jp_journal_magic; /* random value made on fs creation (this | |
803 | * was sb_journal_block_count) */ | |
804 | __le32 jp_journal_max_batch; /* max number of blocks to batch into a | |
805 | * trans */ | |
806 | __le32 jp_journal_max_commit_age; /* in seconds, how old can an async | |
807 | * commit be */ | |
808 | __le32 jp_journal_max_trans_age; /* in seconds, how old can a transaction | |
809 | * be */ | |
810 | }; | |
811 | ||
812 | /* this is the super from 3.5.X, where X >= 10 */ | |
813 | struct reiserfs_super_block_v1 { | |
814 | __le32 s_block_count; /* blocks count */ | |
815 | __le32 s_free_blocks; /* free blocks count */ | |
816 | __le32 s_root_block; /* root block number */ | |
817 | struct journal_params s_journal; | |
818 | __le16 s_blocksize; /* block size */ | |
819 | __le16 s_oid_maxsize; /* max size of object id array, see | |
820 | * get_objectid() commentary */ | |
821 | __le16 s_oid_cursize; /* current size of object id array */ | |
822 | __le16 s_umount_state; /* this is set to 1 when filesystem was | |
823 | * umounted, to 2 - when not */ | |
824 | char s_magic[10]; /* reiserfs magic string indicates that | |
825 | * file system is reiserfs: | |
826 | * "ReIsErFs" or "ReIsEr2Fs" or "ReIsEr3Fs" */ | |
827 | __le16 s_fs_state; /* it is set to used by fsck to mark which | |
828 | * phase of rebuilding is done */ | |
829 | __le32 s_hash_function_code; /* indicate, what hash function is being use | |
830 | * to sort names in a directory*/ | |
831 | __le16 s_tree_height; /* height of disk tree */ | |
832 | __le16 s_bmap_nr; /* amount of bitmap blocks needed to address | |
833 | * each block of file system */ | |
834 | __le16 s_version; /* this field is only reliable on filesystem | |
835 | * with non-standard journal */ | |
836 | __le16 s_reserved_for_journal; /* size in blocks of journal area on main | |
837 | * device, we need to keep after | |
838 | * making fs with non-standard journal */ | |
839 | } __attribute__ ((__packed__)); | |
840 | ||
841 | #define SB_SIZE_V1 (sizeof(struct reiserfs_super_block_v1)) | |
842 | ||
843 | /* this is the on disk super block */ | |
844 | struct reiserfs_super_block { | |
845 | struct reiserfs_super_block_v1 s_v1; | |
846 | __le32 s_inode_generation; | |
847 | __le32 s_flags; /* Right now used only by inode-attributes, if enabled */ | |
848 | unsigned char s_uuid[16]; /* filesystem unique identifier */ | |
849 | unsigned char s_label[16]; /* filesystem volume label */ | |
850 | __le16 s_mnt_count; /* Count of mounts since last fsck */ | |
851 | __le16 s_max_mnt_count; /* Maximum mounts before check */ | |
852 | __le32 s_lastcheck; /* Timestamp of last fsck */ | |
853 | __le32 s_check_interval; /* Interval between checks */ | |
854 | char s_unused[76]; /* zero filled by mkreiserfs and | |
855 | * reiserfs_convert_objectid_map_v1() | |
856 | * so any additions must be updated | |
857 | * there as well. */ | |
858 | } __attribute__ ((__packed__)); | |
859 | ||
860 | #define SB_SIZE (sizeof(struct reiserfs_super_block)) | |
861 | ||
862 | #define REISERFS_VERSION_1 0 | |
863 | #define REISERFS_VERSION_2 2 | |
864 | ||
865 | // on-disk super block fields converted to cpu form | |
866 | #define SB_DISK_SUPER_BLOCK(s) (REISERFS_SB(s)->s_rs) | |
867 | #define SB_V1_DISK_SUPER_BLOCK(s) (&(SB_DISK_SUPER_BLOCK(s)->s_v1)) | |
868 | #define SB_BLOCKSIZE(s) \ | |
869 | le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_blocksize)) | |
870 | #define SB_BLOCK_COUNT(s) \ | |
871 | le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_block_count)) | |
872 | #define SB_FREE_BLOCKS(s) \ | |
873 | le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks)) | |
874 | #define SB_REISERFS_MAGIC(s) \ | |
875 | (SB_V1_DISK_SUPER_BLOCK(s)->s_magic) | |
876 | #define SB_ROOT_BLOCK(s) \ | |
877 | le32_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_root_block)) | |
878 | #define SB_TREE_HEIGHT(s) \ | |
879 | le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height)) | |
880 | #define SB_REISERFS_STATE(s) \ | |
881 | le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state)) | |
882 | #define SB_VERSION(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_version)) | |
883 | #define SB_BMAP_NR(s) le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr)) | |
884 | ||
885 | #define PUT_SB_BLOCK_COUNT(s, val) \ | |
886 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_block_count = cpu_to_le32(val); } while (0) | |
887 | #define PUT_SB_FREE_BLOCKS(s, val) \ | |
888 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_free_blocks = cpu_to_le32(val); } while (0) | |
889 | #define PUT_SB_ROOT_BLOCK(s, val) \ | |
890 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_root_block = cpu_to_le32(val); } while (0) | |
891 | #define PUT_SB_TREE_HEIGHT(s, val) \ | |
892 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_tree_height = cpu_to_le16(val); } while (0) | |
893 | #define PUT_SB_REISERFS_STATE(s, val) \ | |
894 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_umount_state = cpu_to_le16(val); } while (0) | |
895 | #define PUT_SB_VERSION(s, val) \ | |
896 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_version = cpu_to_le16(val); } while (0) | |
897 | #define PUT_SB_BMAP_NR(s, val) \ | |
898 | do { SB_V1_DISK_SUPER_BLOCK(s)->s_bmap_nr = cpu_to_le16 (val); } while (0) | |
899 | ||
900 | #define SB_ONDISK_JP(s) (&SB_V1_DISK_SUPER_BLOCK(s)->s_journal) | |
901 | #define SB_ONDISK_JOURNAL_SIZE(s) \ | |
902 | le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_size)) | |
903 | #define SB_ONDISK_JOURNAL_1st_BLOCK(s) \ | |
904 | le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_1st_block)) | |
905 | #define SB_ONDISK_JOURNAL_DEVICE(s) \ | |
906 | le32_to_cpu ((SB_ONDISK_JP(s)->jp_journal_dev)) | |
907 | #define SB_ONDISK_RESERVED_FOR_JOURNAL(s) \ | |
908 | le16_to_cpu ((SB_V1_DISK_SUPER_BLOCK(s)->s_reserved_for_journal)) | |
909 | ||
910 | #define is_block_in_log_or_reserved_area(s, block) \ | |
911 | block >= SB_JOURNAL_1st_RESERVED_BLOCK(s) \ | |
912 | && block < SB_JOURNAL_1st_RESERVED_BLOCK(s) + \ | |
913 | ((!is_reiserfs_jr(SB_DISK_SUPER_BLOCK(s)) ? \ | |
914 | SB_ONDISK_JOURNAL_SIZE(s) + 1 : SB_ONDISK_RESERVED_FOR_JOURNAL(s))) | |
915 | ||
916 | int is_reiserfs_3_5(struct reiserfs_super_block *rs); | |
917 | int is_reiserfs_3_6(struct reiserfs_super_block *rs); | |
918 | int is_reiserfs_jr(struct reiserfs_super_block *rs); | |
919 | ||
920 | /* ReiserFS leaves the first 64k unused, so that partition labels have | |
921 | enough space. If someone wants to write a fancy bootloader that | |
922 | needs more than 64k, let us know, and this will be increased in size. | |
923 | This number must be larger than than the largest block size on any | |
924 | platform, or code will break. -Hans */ | |
925 | #define REISERFS_DISK_OFFSET_IN_BYTES (64 * 1024) | |
926 | #define REISERFS_FIRST_BLOCK unused_define | |
927 | #define REISERFS_JOURNAL_OFFSET_IN_BYTES REISERFS_DISK_OFFSET_IN_BYTES | |
928 | ||
929 | /* the spot for the super in versions 3.5 - 3.5.10 (inclusive) */ | |
930 | #define REISERFS_OLD_DISK_OFFSET_IN_BYTES (8 * 1024) | |
931 | ||
932 | /* reiserfs internal error code (used by search_by_key and fix_nodes)) */ | |
933 | #define CARRY_ON 0 | |
934 | #define REPEAT_SEARCH -1 | |
935 | #define IO_ERROR -2 | |
936 | #define NO_DISK_SPACE -3 | |
937 | #define NO_BALANCING_NEEDED (-4) | |
938 | #define NO_MORE_UNUSED_CONTIGUOUS_BLOCKS (-5) | |
939 | #define QUOTA_EXCEEDED -6 | |
940 | ||
941 | typedef __u32 b_blocknr_t; | |
942 | typedef __le32 unp_t; | |
943 | ||
944 | struct unfm_nodeinfo { | |
945 | unp_t unfm_nodenum; | |
946 | unsigned short unfm_freespace; | |
947 | }; | |
948 | ||
949 | /* there are two formats of keys: 3.5 and 3.6 | |
950 | */ | |
951 | #define KEY_FORMAT_3_5 0 | |
952 | #define KEY_FORMAT_3_6 1 | |
953 | ||
954 | /* there are two stat datas */ | |
955 | #define STAT_DATA_V1 0 | |
956 | #define STAT_DATA_V2 1 | |
957 | ||
958 | static inline struct reiserfs_inode_info *REISERFS_I(const struct inode *inode) | |
959 | { | |
960 | return container_of(inode, struct reiserfs_inode_info, vfs_inode); | |
961 | } | |
962 | ||
963 | static inline struct reiserfs_sb_info *REISERFS_SB(const struct super_block *sb) | |
964 | { | |
965 | return sb->s_fs_info; | |
966 | } | |
967 | ||
968 | /* Don't trust REISERFS_SB(sb)->s_bmap_nr, it's a u16 | |
969 | * which overflows on large file systems. */ | |
970 | static inline __u32 reiserfs_bmap_count(struct super_block *sb) | |
971 | { | |
972 | return (SB_BLOCK_COUNT(sb) - 1) / (sb->s_blocksize * 8) + 1; | |
973 | } | |
974 | ||
975 | static inline int bmap_would_wrap(unsigned bmap_nr) | |
976 | { | |
977 | return bmap_nr > ((1LL << 16) - 1); | |
978 | } | |
979 | ||
980 | /** this says about version of key of all items (but stat data) the | |
981 | object consists of */ | |
982 | #define get_inode_item_key_version( inode ) \ | |
983 | ((REISERFS_I(inode)->i_flags & i_item_key_version_mask) ? KEY_FORMAT_3_6 : KEY_FORMAT_3_5) | |
984 | ||
985 | #define set_inode_item_key_version( inode, version ) \ | |
986 | ({ if((version)==KEY_FORMAT_3_6) \ | |
987 | REISERFS_I(inode)->i_flags |= i_item_key_version_mask; \ | |
988 | else \ | |
989 | REISERFS_I(inode)->i_flags &= ~i_item_key_version_mask; }) | |
990 | ||
991 | #define get_inode_sd_version(inode) \ | |
992 | ((REISERFS_I(inode)->i_flags & i_stat_data_version_mask) ? STAT_DATA_V2 : STAT_DATA_V1) | |
993 | ||
994 | #define set_inode_sd_version(inode, version) \ | |
995 | ({ if((version)==STAT_DATA_V2) \ | |
996 | REISERFS_I(inode)->i_flags |= i_stat_data_version_mask; \ | |
997 | else \ | |
998 | REISERFS_I(inode)->i_flags &= ~i_stat_data_version_mask; }) | |
999 | ||
1000 | /* This is an aggressive tail suppression policy, I am hoping it | |
1001 | improves our benchmarks. The principle behind it is that percentage | |
1002 | space saving is what matters, not absolute space saving. This is | |
1003 | non-intuitive, but it helps to understand it if you consider that the | |
1004 | cost to access 4 blocks is not much more than the cost to access 1 | |
1005 | block, if you have to do a seek and rotate. A tail risks a | |
1006 | non-linear disk access that is significant as a percentage of total | |
1007 | time cost for a 4 block file and saves an amount of space that is | |
1008 | less significant as a percentage of space, or so goes the hypothesis. | |
1009 | -Hans */ | |
1010 | #define STORE_TAIL_IN_UNFM_S1(n_file_size,n_tail_size,n_block_size) \ | |
1011 | (\ | |
1012 | (!(n_tail_size)) || \ | |
1013 | (((n_tail_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) || \ | |
1014 | ( (n_file_size) >= (n_block_size) * 4 ) || \ | |
1015 | ( ( (n_file_size) >= (n_block_size) * 3 ) && \ | |
1016 | ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/4) ) || \ | |
1017 | ( ( (n_file_size) >= (n_block_size) * 2 ) && \ | |
1018 | ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size))/2) ) || \ | |
1019 | ( ( (n_file_size) >= (n_block_size) ) && \ | |
1020 | ( (n_tail_size) >= (MAX_DIRECT_ITEM_LEN(n_block_size) * 3)/4) ) ) \ | |
1021 | ) | |
1022 | ||
1023 | /* Another strategy for tails, this one means only create a tail if all the | |
1024 | file would fit into one DIRECT item. | |
1025 | Primary intention for this one is to increase performance by decreasing | |
1026 | seeking. | |
1027 | */ | |
1028 | #define STORE_TAIL_IN_UNFM_S2(n_file_size,n_tail_size,n_block_size) \ | |
1029 | (\ | |
1030 | (!(n_tail_size)) || \ | |
1031 | (((n_file_size) > MAX_DIRECT_ITEM_LEN(n_block_size)) ) \ | |
1032 | ) | |
1033 | ||
1034 | /* | |
1035 | * values for s_umount_state field | |
1036 | */ | |
1037 | #define REISERFS_VALID_FS 1 | |
1038 | #define REISERFS_ERROR_FS 2 | |
1039 | ||
1040 | // | |
1041 | // there are 5 item types currently | |
1042 | // | |
1043 | #define TYPE_STAT_DATA 0 | |
1044 | #define TYPE_INDIRECT 1 | |
1045 | #define TYPE_DIRECT 2 | |
1046 | #define TYPE_DIRENTRY 3 | |
1047 | #define TYPE_MAXTYPE 3 | |
1048 | #define TYPE_ANY 15 // FIXME: comment is required | |
1049 | ||
1050 | /***************************************************************************/ | |
1051 | /* KEY & ITEM HEAD */ | |
1052 | /***************************************************************************/ | |
1053 | ||
1054 | // | |
1055 | // directories use this key as well as old files | |
1056 | // | |
1057 | struct offset_v1 { | |
1058 | __le32 k_offset; | |
1059 | __le32 k_uniqueness; | |
1060 | } __attribute__ ((__packed__)); | |
1061 | ||
1062 | struct offset_v2 { | |
1063 | __le64 v; | |
1064 | } __attribute__ ((__packed__)); | |
1065 | ||
1066 | static inline __u16 offset_v2_k_type(const struct offset_v2 *v2) | |
1067 | { | |
1068 | __u8 type = le64_to_cpu(v2->v) >> 60; | |
1069 | return (type <= TYPE_MAXTYPE) ? type : TYPE_ANY; | |
1070 | } | |
1071 | ||
1072 | static inline void set_offset_v2_k_type(struct offset_v2 *v2, int type) | |
1073 | { | |
1074 | v2->v = | |
1075 | (v2->v & cpu_to_le64(~0ULL >> 4)) | cpu_to_le64((__u64) type << 60); | |
1076 | } | |
1077 | ||
1078 | static inline loff_t offset_v2_k_offset(const struct offset_v2 *v2) | |
1079 | { | |
1080 | return le64_to_cpu(v2->v) & (~0ULL >> 4); | |
1081 | } | |
1082 | ||
1083 | static inline void set_offset_v2_k_offset(struct offset_v2 *v2, loff_t offset) | |
1084 | { | |
1085 | offset &= (~0ULL >> 4); | |
1086 | v2->v = (v2->v & cpu_to_le64(15ULL << 60)) | cpu_to_le64(offset); | |
1087 | } | |
1088 | ||
1089 | /* Key of an item determines its location in the S+tree, and | |
1090 | is composed of 4 components */ | |
1091 | struct reiserfs_key { | |
1092 | __le32 k_dir_id; /* packing locality: by default parent | |
1093 | directory object id */ | |
1094 | __le32 k_objectid; /* object identifier */ | |
1095 | union { | |
1096 | struct offset_v1 k_offset_v1; | |
1097 | struct offset_v2 k_offset_v2; | |
1098 | } __attribute__ ((__packed__)) u; | |
1099 | } __attribute__ ((__packed__)); | |
1100 | ||
1101 | struct in_core_key { | |
1102 | __u32 k_dir_id; /* packing locality: by default parent | |
1103 | directory object id */ | |
1104 | __u32 k_objectid; /* object identifier */ | |
1105 | __u64 k_offset; | |
1106 | __u8 k_type; | |
1107 | }; | |
1108 | ||
1109 | struct cpu_key { | |
1110 | struct in_core_key on_disk_key; | |
1111 | int version; | |
1112 | int key_length; /* 3 in all cases but direct2indirect and | |
1113 | indirect2direct conversion */ | |
1114 | }; | |
1115 | ||
1116 | /* Our function for comparing keys can compare keys of different | |
1117 | lengths. It takes as a parameter the length of the keys it is to | |
1118 | compare. These defines are used in determining what is to be passed | |
1119 | to it as that parameter. */ | |
1120 | #define REISERFS_FULL_KEY_LEN 4 | |
1121 | #define REISERFS_SHORT_KEY_LEN 2 | |
1122 | ||
1123 | /* The result of the key compare */ | |
1124 | #define FIRST_GREATER 1 | |
1125 | #define SECOND_GREATER -1 | |
1126 | #define KEYS_IDENTICAL 0 | |
1127 | #define KEY_FOUND 1 | |
1128 | #define KEY_NOT_FOUND 0 | |
1129 | ||
1130 | #define KEY_SIZE (sizeof(struct reiserfs_key)) | |
1131 | #define SHORT_KEY_SIZE (sizeof (__u32) + sizeof (__u32)) | |
1132 | ||
1133 | /* return values for search_by_key and clones */ | |
1134 | #define ITEM_FOUND 1 | |
1135 | #define ITEM_NOT_FOUND 0 | |
1136 | #define ENTRY_FOUND 1 | |
1137 | #define ENTRY_NOT_FOUND 0 | |
1138 | #define DIRECTORY_NOT_FOUND -1 | |
1139 | #define REGULAR_FILE_FOUND -2 | |
1140 | #define DIRECTORY_FOUND -3 | |
1141 | #define BYTE_FOUND 1 | |
1142 | #define BYTE_NOT_FOUND 0 | |
1143 | #define FILE_NOT_FOUND -1 | |
1144 | ||
1145 | #define POSITION_FOUND 1 | |
1146 | #define POSITION_NOT_FOUND 0 | |
1147 | ||
1148 | // return values for reiserfs_find_entry and search_by_entry_key | |
1149 | #define NAME_FOUND 1 | |
1150 | #define NAME_NOT_FOUND 0 | |
1151 | #define GOTO_PREVIOUS_ITEM 2 | |
1152 | #define NAME_FOUND_INVISIBLE 3 | |
1153 | ||
1154 | /* Everything in the filesystem is stored as a set of items. The | |
1155 | item head contains the key of the item, its free space (for | |
1156 | indirect items) and specifies the location of the item itself | |
1157 | within the block. */ | |
1158 | ||
1159 | struct item_head { | |
1160 | /* Everything in the tree is found by searching for it based on | |
1161 | * its key.*/ | |
1162 | struct reiserfs_key ih_key; | |
1163 | union { | |
1164 | /* The free space in the last unformatted node of an | |
1165 | indirect item if this is an indirect item. This | |
1166 | equals 0xFFFF iff this is a direct item or stat data | |
1167 | item. Note that the key, not this field, is used to | |
1168 | determine the item type, and thus which field this | |
1169 | union contains. */ | |
1170 | __le16 ih_free_space_reserved; | |
1171 | /* Iff this is a directory item, this field equals the | |
1172 | number of directory entries in the directory item. */ | |
1173 | __le16 ih_entry_count; | |
1174 | } __attribute__ ((__packed__)) u; | |
1175 | __le16 ih_item_len; /* total size of the item body */ | |
1176 | __le16 ih_item_location; /* an offset to the item body | |
1177 | * within the block */ | |
1178 | __le16 ih_version; /* 0 for all old items, 2 for new | |
1179 | ones. Highest bit is set by fsck | |
1180 | temporary, cleaned after all | |
1181 | done */ | |
1182 | } __attribute__ ((__packed__)); | |
1183 | /* size of item header */ | |
1184 | #define IH_SIZE (sizeof(struct item_head)) | |
1185 | ||
1186 | #define ih_free_space(ih) le16_to_cpu((ih)->u.ih_free_space_reserved) | |
1187 | #define ih_version(ih) le16_to_cpu((ih)->ih_version) | |
1188 | #define ih_entry_count(ih) le16_to_cpu((ih)->u.ih_entry_count) | |
1189 | #define ih_location(ih) le16_to_cpu((ih)->ih_item_location) | |
1190 | #define ih_item_len(ih) le16_to_cpu((ih)->ih_item_len) | |
1191 | ||
1192 | #define put_ih_free_space(ih, val) do { (ih)->u.ih_free_space_reserved = cpu_to_le16(val); } while(0) | |
1193 | #define put_ih_version(ih, val) do { (ih)->ih_version = cpu_to_le16(val); } while (0) | |
1194 | #define put_ih_entry_count(ih, val) do { (ih)->u.ih_entry_count = cpu_to_le16(val); } while (0) | |
1195 | #define put_ih_location(ih, val) do { (ih)->ih_item_location = cpu_to_le16(val); } while (0) | |
1196 | #define put_ih_item_len(ih, val) do { (ih)->ih_item_len = cpu_to_le16(val); } while (0) | |
1197 | ||
1198 | #define unreachable_item(ih) (ih_version(ih) & (1 << 15)) | |
1199 | ||
1200 | #define get_ih_free_space(ih) (ih_version (ih) == KEY_FORMAT_3_6 ? 0 : ih_free_space (ih)) | |
1201 | #define set_ih_free_space(ih,val) put_ih_free_space((ih), ((ih_version(ih) == KEY_FORMAT_3_6) ? 0 : (val))) | |
1202 | ||
1203 | /* these operate on indirect items, where you've got an array of ints | |
1204 | ** at a possibly unaligned location. These are a noop on ia32 | |
1205 | ** | |
1206 | ** p is the array of __u32, i is the index into the array, v is the value | |
1207 | ** to store there. | |
1208 | */ | |
1209 | #define get_block_num(p, i) get_unaligned_le32((p) + (i)) | |
1210 | #define put_block_num(p, i, v) put_unaligned_le32((v), (p) + (i)) | |
1211 | ||
1212 | // | |
1213 | // in old version uniqueness field shows key type | |
1214 | // | |
1215 | #define V1_SD_UNIQUENESS 0 | |
1216 | #define V1_INDIRECT_UNIQUENESS 0xfffffffe | |
1217 | #define V1_DIRECT_UNIQUENESS 0xffffffff | |
1218 | #define V1_DIRENTRY_UNIQUENESS 500 | |
1219 | #define V1_ANY_UNIQUENESS 555 // FIXME: comment is required | |
1220 | ||
1221 | // | |
1222 | // here are conversion routines | |
1223 | // | |
1224 | static inline int uniqueness2type(__u32 uniqueness) CONSTF; | |
1225 | static inline int uniqueness2type(__u32 uniqueness) | |
1226 | { | |
1227 | switch ((int)uniqueness) { | |
1228 | case V1_SD_UNIQUENESS: | |
1229 | return TYPE_STAT_DATA; | |
1230 | case V1_INDIRECT_UNIQUENESS: | |
1231 | return TYPE_INDIRECT; | |
1232 | case V1_DIRECT_UNIQUENESS: | |
1233 | return TYPE_DIRECT; | |
1234 | case V1_DIRENTRY_UNIQUENESS: | |
1235 | return TYPE_DIRENTRY; | |
1236 | case V1_ANY_UNIQUENESS: | |
1237 | default: | |
1238 | return TYPE_ANY; | |
1239 | } | |
1240 | } | |
1241 | ||
1242 | static inline __u32 type2uniqueness(int type) CONSTF; | |
1243 | static inline __u32 type2uniqueness(int type) | |
1244 | { | |
1245 | switch (type) { | |
1246 | case TYPE_STAT_DATA: | |
1247 | return V1_SD_UNIQUENESS; | |
1248 | case TYPE_INDIRECT: | |
1249 | return V1_INDIRECT_UNIQUENESS; | |
1250 | case TYPE_DIRECT: | |
1251 | return V1_DIRECT_UNIQUENESS; | |
1252 | case TYPE_DIRENTRY: | |
1253 | return V1_DIRENTRY_UNIQUENESS; | |
1254 | case TYPE_ANY: | |
1255 | default: | |
1256 | return V1_ANY_UNIQUENESS; | |
1257 | } | |
1258 | } | |
1259 | ||
1260 | // | |
1261 | // key is pointer to on disk key which is stored in le, result is cpu, | |
1262 | // there is no way to get version of object from key, so, provide | |
1263 | // version to these defines | |
1264 | // | |
1265 | static inline loff_t le_key_k_offset(int version, | |
1266 | const struct reiserfs_key *key) | |
1267 | { | |
1268 | return (version == KEY_FORMAT_3_5) ? | |
1269 | le32_to_cpu(key->u.k_offset_v1.k_offset) : | |
1270 | offset_v2_k_offset(&(key->u.k_offset_v2)); | |
1271 | } | |
1272 | ||
1273 | static inline loff_t le_ih_k_offset(const struct item_head *ih) | |
1274 | { | |
1275 | return le_key_k_offset(ih_version(ih), &(ih->ih_key)); | |
1276 | } | |
1277 | ||
1278 | static inline loff_t le_key_k_type(int version, const struct reiserfs_key *key) | |
1279 | { | |
4cf5f7ad JM |
1280 | if (version == KEY_FORMAT_3_5) { |
1281 | loff_t val = le32_to_cpu(key->u.k_offset_v1.k_uniqueness); | |
1282 | return uniqueness2type(val); | |
1283 | } else | |
1284 | return offset_v2_k_type(&(key->u.k_offset_v2)); | |
f466c6fd AV |
1285 | } |
1286 | ||
1287 | static inline loff_t le_ih_k_type(const struct item_head *ih) | |
1288 | { | |
1289 | return le_key_k_type(ih_version(ih), &(ih->ih_key)); | |
1290 | } | |
1291 | ||
1292 | static inline void set_le_key_k_offset(int version, struct reiserfs_key *key, | |
1293 | loff_t offset) | |
1294 | { | |
4cf5f7ad JM |
1295 | if (version == KEY_FORMAT_3_5) |
1296 | key->u.k_offset_v1.k_offset = cpu_to_le32(offset); | |
1297 | else | |
1298 | set_offset_v2_k_offset(&key->u.k_offset_v2, offset); | |
1299 | } | |
1300 | ||
1301 | static inline void add_le_key_k_offset(int version, struct reiserfs_key *key, | |
1302 | loff_t offset) | |
1303 | { | |
1304 | set_le_key_k_offset(version, key, | |
1305 | le_key_k_offset(version, key) + offset); | |
1306 | } | |
1307 | ||
1308 | static inline void add_le_ih_k_offset(struct item_head *ih, loff_t offset) | |
1309 | { | |
1310 | add_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset); | |
f466c6fd AV |
1311 | } |
1312 | ||
1313 | static inline void set_le_ih_k_offset(struct item_head *ih, loff_t offset) | |
1314 | { | |
1315 | set_le_key_k_offset(ih_version(ih), &(ih->ih_key), offset); | |
1316 | } | |
1317 | ||
1318 | static inline void set_le_key_k_type(int version, struct reiserfs_key *key, | |
1319 | int type) | |
1320 | { | |
4cf5f7ad JM |
1321 | if (version == KEY_FORMAT_3_5) { |
1322 | type = type2uniqueness(type); | |
1323 | key->u.k_offset_v1.k_uniqueness = cpu_to_le32(type); | |
1324 | } else | |
1325 | set_offset_v2_k_type(&key->u.k_offset_v2, type); | |
f466c6fd AV |
1326 | } |
1327 | ||
1328 | static inline void set_le_ih_k_type(struct item_head *ih, int type) | |
1329 | { | |
1330 | set_le_key_k_type(ih_version(ih), &(ih->ih_key), type); | |
1331 | } | |
1332 | ||
1333 | static inline int is_direntry_le_key(int version, struct reiserfs_key *key) | |
1334 | { | |
1335 | return le_key_k_type(version, key) == TYPE_DIRENTRY; | |
1336 | } | |
1337 | ||
1338 | static inline int is_direct_le_key(int version, struct reiserfs_key *key) | |
1339 | { | |
1340 | return le_key_k_type(version, key) == TYPE_DIRECT; | |
1341 | } | |
1342 | ||
1343 | static inline int is_indirect_le_key(int version, struct reiserfs_key *key) | |
1344 | { | |
1345 | return le_key_k_type(version, key) == TYPE_INDIRECT; | |
1346 | } | |
1347 | ||
1348 | static inline int is_statdata_le_key(int version, struct reiserfs_key *key) | |
1349 | { | |
1350 | return le_key_k_type(version, key) == TYPE_STAT_DATA; | |
1351 | } | |
1352 | ||
1353 | // | |
1354 | // item header has version. | |
1355 | // | |
1356 | static inline int is_direntry_le_ih(struct item_head *ih) | |
1357 | { | |
1358 | return is_direntry_le_key(ih_version(ih), &ih->ih_key); | |
1359 | } | |
1360 | ||
1361 | static inline int is_direct_le_ih(struct item_head *ih) | |
1362 | { | |
1363 | return is_direct_le_key(ih_version(ih), &ih->ih_key); | |
1364 | } | |
1365 | ||
1366 | static inline int is_indirect_le_ih(struct item_head *ih) | |
1367 | { | |
1368 | return is_indirect_le_key(ih_version(ih), &ih->ih_key); | |
1369 | } | |
1370 | ||
1371 | static inline int is_statdata_le_ih(struct item_head *ih) | |
1372 | { | |
1373 | return is_statdata_le_key(ih_version(ih), &ih->ih_key); | |
1374 | } | |
1375 | ||
1376 | // | |
1377 | // key is pointer to cpu key, result is cpu | |
1378 | // | |
1379 | static inline loff_t cpu_key_k_offset(const struct cpu_key *key) | |
1380 | { | |
1381 | return key->on_disk_key.k_offset; | |
1382 | } | |
1383 | ||
1384 | static inline loff_t cpu_key_k_type(const struct cpu_key *key) | |
1385 | { | |
1386 | return key->on_disk_key.k_type; | |
1387 | } | |
1388 | ||
1389 | static inline void set_cpu_key_k_offset(struct cpu_key *key, loff_t offset) | |
1390 | { | |
1391 | key->on_disk_key.k_offset = offset; | |
1392 | } | |
1393 | ||
1394 | static inline void set_cpu_key_k_type(struct cpu_key *key, int type) | |
1395 | { | |
1396 | key->on_disk_key.k_type = type; | |
1397 | } | |
1398 | ||
1399 | static inline void cpu_key_k_offset_dec(struct cpu_key *key) | |
1400 | { | |
1401 | key->on_disk_key.k_offset--; | |
1402 | } | |
1403 | ||
1404 | #define is_direntry_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRENTRY) | |
1405 | #define is_direct_cpu_key(key) (cpu_key_k_type (key) == TYPE_DIRECT) | |
1406 | #define is_indirect_cpu_key(key) (cpu_key_k_type (key) == TYPE_INDIRECT) | |
1407 | #define is_statdata_cpu_key(key) (cpu_key_k_type (key) == TYPE_STAT_DATA) | |
1408 | ||
1409 | /* are these used ? */ | |
1410 | #define is_direntry_cpu_ih(ih) (is_direntry_cpu_key (&((ih)->ih_key))) | |
1411 | #define is_direct_cpu_ih(ih) (is_direct_cpu_key (&((ih)->ih_key))) | |
1412 | #define is_indirect_cpu_ih(ih) (is_indirect_cpu_key (&((ih)->ih_key))) | |
1413 | #define is_statdata_cpu_ih(ih) (is_statdata_cpu_key (&((ih)->ih_key))) | |
1414 | ||
1415 | #define I_K_KEY_IN_ITEM(ih, key, n_blocksize) \ | |
1416 | (!COMP_SHORT_KEYS(ih, key) && \ | |
1417 | I_OFF_BYTE_IN_ITEM(ih, k_offset(key), n_blocksize)) | |
1418 | ||
1419 | /* maximal length of item */ | |
1420 | #define MAX_ITEM_LEN(block_size) (block_size - BLKH_SIZE - IH_SIZE) | |
1421 | #define MIN_ITEM_LEN 1 | |
1422 | ||
1423 | /* object identifier for root dir */ | |
1424 | #define REISERFS_ROOT_OBJECTID 2 | |
1425 | #define REISERFS_ROOT_PARENT_OBJECTID 1 | |
1426 | ||
1427 | extern struct reiserfs_key root_key; | |
1428 | ||
1429 | /* | |
1430 | * Picture represents a leaf of the S+tree | |
1431 | * ______________________________________________________ | |
1432 | * | | Array of | | | | |
1433 | * |Block | Object-Item | F r e e | Objects- | | |
1434 | * | head | Headers | S p a c e | Items | | |
1435 | * |______|_______________|___________________|___________| | |
1436 | */ | |
1437 | ||
1438 | /* Header of a disk block. More precisely, header of a formatted leaf | |
1439 | or internal node, and not the header of an unformatted node. */ | |
1440 | struct block_head { | |
1441 | __le16 blk_level; /* Level of a block in the tree. */ | |
1442 | __le16 blk_nr_item; /* Number of keys/items in a block. */ | |
1443 | __le16 blk_free_space; /* Block free space in bytes. */ | |
1444 | __le16 blk_reserved; | |
1445 | /* dump this in v4/planA */ | |
1446 | struct reiserfs_key blk_right_delim_key; /* kept only for compatibility */ | |
1447 | }; | |
1448 | ||
1449 | #define BLKH_SIZE (sizeof(struct block_head)) | |
1450 | #define blkh_level(p_blkh) (le16_to_cpu((p_blkh)->blk_level)) | |
1451 | #define blkh_nr_item(p_blkh) (le16_to_cpu((p_blkh)->blk_nr_item)) | |
1452 | #define blkh_free_space(p_blkh) (le16_to_cpu((p_blkh)->blk_free_space)) | |
1453 | #define blkh_reserved(p_blkh) (le16_to_cpu((p_blkh)->blk_reserved)) | |
1454 | #define set_blkh_level(p_blkh,val) ((p_blkh)->blk_level = cpu_to_le16(val)) | |
1455 | #define set_blkh_nr_item(p_blkh,val) ((p_blkh)->blk_nr_item = cpu_to_le16(val)) | |
1456 | #define set_blkh_free_space(p_blkh,val) ((p_blkh)->blk_free_space = cpu_to_le16(val)) | |
1457 | #define set_blkh_reserved(p_blkh,val) ((p_blkh)->blk_reserved = cpu_to_le16(val)) | |
1458 | #define blkh_right_delim_key(p_blkh) ((p_blkh)->blk_right_delim_key) | |
1459 | #define set_blkh_right_delim_key(p_blkh,val) ((p_blkh)->blk_right_delim_key = val) | |
1460 | ||
1461 | /* | |
1462 | * values for blk_level field of the struct block_head | |
1463 | */ | |
1464 | ||
1465 | #define FREE_LEVEL 0 /* when node gets removed from the tree its | |
1466 | blk_level is set to FREE_LEVEL. It is then | |
1467 | used to see whether the node is still in the | |
1468 | tree */ | |
1469 | ||
1470 | #define DISK_LEAF_NODE_LEVEL 1 /* Leaf node level. */ | |
1471 | ||
1472 | /* Given the buffer head of a formatted node, resolve to the block head of that node. */ | |
1473 | #define B_BLK_HEAD(bh) ((struct block_head *)((bh)->b_data)) | |
1474 | /* Number of items that are in buffer. */ | |
1475 | #define B_NR_ITEMS(bh) (blkh_nr_item(B_BLK_HEAD(bh))) | |
1476 | #define B_LEVEL(bh) (blkh_level(B_BLK_HEAD(bh))) | |
1477 | #define B_FREE_SPACE(bh) (blkh_free_space(B_BLK_HEAD(bh))) | |
1478 | ||
1479 | #define PUT_B_NR_ITEMS(bh, val) do { set_blkh_nr_item(B_BLK_HEAD(bh), val); } while (0) | |
1480 | #define PUT_B_LEVEL(bh, val) do { set_blkh_level(B_BLK_HEAD(bh), val); } while (0) | |
1481 | #define PUT_B_FREE_SPACE(bh, val) do { set_blkh_free_space(B_BLK_HEAD(bh), val); } while (0) | |
1482 | ||
1483 | /* Get right delimiting key. -- little endian */ | |
1484 | #define B_PRIGHT_DELIM_KEY(bh) (&(blk_right_delim_key(B_BLK_HEAD(bh)))) | |
1485 | ||
1486 | /* Does the buffer contain a disk leaf. */ | |
1487 | #define B_IS_ITEMS_LEVEL(bh) (B_LEVEL(bh) == DISK_LEAF_NODE_LEVEL) | |
1488 | ||
1489 | /* Does the buffer contain a disk internal node */ | |
1490 | #define B_IS_KEYS_LEVEL(bh) (B_LEVEL(bh) > DISK_LEAF_NODE_LEVEL \ | |
1491 | && B_LEVEL(bh) <= MAX_HEIGHT) | |
1492 | ||
1493 | /***************************************************************************/ | |
1494 | /* STAT DATA */ | |
1495 | /***************************************************************************/ | |
1496 | ||
1497 | // | |
1498 | // old stat data is 32 bytes long. We are going to distinguish new one by | |
1499 | // different size | |
1500 | // | |
1501 | struct stat_data_v1 { | |
1502 | __le16 sd_mode; /* file type, permissions */ | |
1503 | __le16 sd_nlink; /* number of hard links */ | |
1504 | __le16 sd_uid; /* owner */ | |
1505 | __le16 sd_gid; /* group */ | |
1506 | __le32 sd_size; /* file size */ | |
1507 | __le32 sd_atime; /* time of last access */ | |
1508 | __le32 sd_mtime; /* time file was last modified */ | |
1509 | __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */ | |
1510 | union { | |
1511 | __le32 sd_rdev; | |
1512 | __le32 sd_blocks; /* number of blocks file uses */ | |
1513 | } __attribute__ ((__packed__)) u; | |
1514 | __le32 sd_first_direct_byte; /* first byte of file which is stored | |
1515 | in a direct item: except that if it | |
1516 | equals 1 it is a symlink and if it | |
1517 | equals ~(__u32)0 there is no | |
1518 | direct item. The existence of this | |
1519 | field really grates on me. Let's | |
1520 | replace it with a macro based on | |
1521 | sd_size and our tail suppression | |
1522 | policy. Someday. -Hans */ | |
1523 | } __attribute__ ((__packed__)); | |
1524 | ||
1525 | #define SD_V1_SIZE (sizeof(struct stat_data_v1)) | |
1526 | #define stat_data_v1(ih) (ih_version (ih) == KEY_FORMAT_3_5) | |
1527 | #define sd_v1_mode(sdp) (le16_to_cpu((sdp)->sd_mode)) | |
1528 | #define set_sd_v1_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v)) | |
1529 | #define sd_v1_nlink(sdp) (le16_to_cpu((sdp)->sd_nlink)) | |
1530 | #define set_sd_v1_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le16(v)) | |
1531 | #define sd_v1_uid(sdp) (le16_to_cpu((sdp)->sd_uid)) | |
1532 | #define set_sd_v1_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le16(v)) | |
1533 | #define sd_v1_gid(sdp) (le16_to_cpu((sdp)->sd_gid)) | |
1534 | #define set_sd_v1_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le16(v)) | |
1535 | #define sd_v1_size(sdp) (le32_to_cpu((sdp)->sd_size)) | |
1536 | #define set_sd_v1_size(sdp,v) ((sdp)->sd_size = cpu_to_le32(v)) | |
1537 | #define sd_v1_atime(sdp) (le32_to_cpu((sdp)->sd_atime)) | |
1538 | #define set_sd_v1_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v)) | |
1539 | #define sd_v1_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime)) | |
1540 | #define set_sd_v1_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v)) | |
1541 | #define sd_v1_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime)) | |
1542 | #define set_sd_v1_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v)) | |
1543 | #define sd_v1_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev)) | |
1544 | #define set_sd_v1_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v)) | |
1545 | #define sd_v1_blocks(sdp) (le32_to_cpu((sdp)->u.sd_blocks)) | |
1546 | #define set_sd_v1_blocks(sdp,v) ((sdp)->u.sd_blocks = cpu_to_le32(v)) | |
1547 | #define sd_v1_first_direct_byte(sdp) \ | |
1548 | (le32_to_cpu((sdp)->sd_first_direct_byte)) | |
1549 | #define set_sd_v1_first_direct_byte(sdp,v) \ | |
1550 | ((sdp)->sd_first_direct_byte = cpu_to_le32(v)) | |
1551 | ||
1552 | /* inode flags stored in sd_attrs (nee sd_reserved) */ | |
1553 | ||
1554 | /* we want common flags to have the same values as in ext2, | |
1555 | so chattr(1) will work without problems */ | |
1556 | #define REISERFS_IMMUTABLE_FL FS_IMMUTABLE_FL | |
1557 | #define REISERFS_APPEND_FL FS_APPEND_FL | |
1558 | #define REISERFS_SYNC_FL FS_SYNC_FL | |
1559 | #define REISERFS_NOATIME_FL FS_NOATIME_FL | |
1560 | #define REISERFS_NODUMP_FL FS_NODUMP_FL | |
1561 | #define REISERFS_SECRM_FL FS_SECRM_FL | |
1562 | #define REISERFS_UNRM_FL FS_UNRM_FL | |
1563 | #define REISERFS_COMPR_FL FS_COMPR_FL | |
1564 | #define REISERFS_NOTAIL_FL FS_NOTAIL_FL | |
1565 | ||
1566 | /* persistent flags that file inherits from the parent directory */ | |
1567 | #define REISERFS_INHERIT_MASK ( REISERFS_IMMUTABLE_FL | \ | |
1568 | REISERFS_SYNC_FL | \ | |
1569 | REISERFS_NOATIME_FL | \ | |
1570 | REISERFS_NODUMP_FL | \ | |
1571 | REISERFS_SECRM_FL | \ | |
1572 | REISERFS_COMPR_FL | \ | |
1573 | REISERFS_NOTAIL_FL ) | |
1574 | ||
1575 | /* Stat Data on disk (reiserfs version of UFS disk inode minus the | |
1576 | address blocks) */ | |
1577 | struct stat_data { | |
1578 | __le16 sd_mode; /* file type, permissions */ | |
1579 | __le16 sd_attrs; /* persistent inode flags */ | |
1580 | __le32 sd_nlink; /* number of hard links */ | |
1581 | __le64 sd_size; /* file size */ | |
1582 | __le32 sd_uid; /* owner */ | |
1583 | __le32 sd_gid; /* group */ | |
1584 | __le32 sd_atime; /* time of last access */ | |
1585 | __le32 sd_mtime; /* time file was last modified */ | |
1586 | __le32 sd_ctime; /* time inode (stat data) was last changed (except changes to sd_atime and sd_mtime) */ | |
1587 | __le32 sd_blocks; | |
1588 | union { | |
1589 | __le32 sd_rdev; | |
1590 | __le32 sd_generation; | |
1591 | //__le32 sd_first_direct_byte; | |
1592 | /* first byte of file which is stored in a | |
1593 | direct item: except that if it equals 1 | |
1594 | it is a symlink and if it equals | |
1595 | ~(__u32)0 there is no direct item. The | |
1596 | existence of this field really grates | |
1597 | on me. Let's replace it with a macro | |
1598 | based on sd_size and our tail | |
1599 | suppression policy? */ | |
1600 | } __attribute__ ((__packed__)) u; | |
1601 | } __attribute__ ((__packed__)); | |
1602 | // | |
1603 | // this is 44 bytes long | |
1604 | // | |
1605 | #define SD_SIZE (sizeof(struct stat_data)) | |
1606 | #define SD_V2_SIZE SD_SIZE | |
1607 | #define stat_data_v2(ih) (ih_version (ih) == KEY_FORMAT_3_6) | |
1608 | #define sd_v2_mode(sdp) (le16_to_cpu((sdp)->sd_mode)) | |
1609 | #define set_sd_v2_mode(sdp,v) ((sdp)->sd_mode = cpu_to_le16(v)) | |
1610 | /* sd_reserved */ | |
1611 | /* set_sd_reserved */ | |
1612 | #define sd_v2_nlink(sdp) (le32_to_cpu((sdp)->sd_nlink)) | |
1613 | #define set_sd_v2_nlink(sdp,v) ((sdp)->sd_nlink = cpu_to_le32(v)) | |
1614 | #define sd_v2_size(sdp) (le64_to_cpu((sdp)->sd_size)) | |
1615 | #define set_sd_v2_size(sdp,v) ((sdp)->sd_size = cpu_to_le64(v)) | |
1616 | #define sd_v2_uid(sdp) (le32_to_cpu((sdp)->sd_uid)) | |
1617 | #define set_sd_v2_uid(sdp,v) ((sdp)->sd_uid = cpu_to_le32(v)) | |
1618 | #define sd_v2_gid(sdp) (le32_to_cpu((sdp)->sd_gid)) | |
1619 | #define set_sd_v2_gid(sdp,v) ((sdp)->sd_gid = cpu_to_le32(v)) | |
1620 | #define sd_v2_atime(sdp) (le32_to_cpu((sdp)->sd_atime)) | |
1621 | #define set_sd_v2_atime(sdp,v) ((sdp)->sd_atime = cpu_to_le32(v)) | |
1622 | #define sd_v2_mtime(sdp) (le32_to_cpu((sdp)->sd_mtime)) | |
1623 | #define set_sd_v2_mtime(sdp,v) ((sdp)->sd_mtime = cpu_to_le32(v)) | |
1624 | #define sd_v2_ctime(sdp) (le32_to_cpu((sdp)->sd_ctime)) | |
1625 | #define set_sd_v2_ctime(sdp,v) ((sdp)->sd_ctime = cpu_to_le32(v)) | |
1626 | #define sd_v2_blocks(sdp) (le32_to_cpu((sdp)->sd_blocks)) | |
1627 | #define set_sd_v2_blocks(sdp,v) ((sdp)->sd_blocks = cpu_to_le32(v)) | |
1628 | #define sd_v2_rdev(sdp) (le32_to_cpu((sdp)->u.sd_rdev)) | |
1629 | #define set_sd_v2_rdev(sdp,v) ((sdp)->u.sd_rdev = cpu_to_le32(v)) | |
1630 | #define sd_v2_generation(sdp) (le32_to_cpu((sdp)->u.sd_generation)) | |
1631 | #define set_sd_v2_generation(sdp,v) ((sdp)->u.sd_generation = cpu_to_le32(v)) | |
1632 | #define sd_v2_attrs(sdp) (le16_to_cpu((sdp)->sd_attrs)) | |
1633 | #define set_sd_v2_attrs(sdp,v) ((sdp)->sd_attrs = cpu_to_le16(v)) | |
1634 | ||
1635 | /***************************************************************************/ | |
1636 | /* DIRECTORY STRUCTURE */ | |
1637 | /***************************************************************************/ | |
1638 | /* | |
1639 | Picture represents the structure of directory items | |
1640 | ________________________________________________ | |
1641 | | Array of | | | | | | | |
1642 | | directory |N-1| N-2 | .... | 1st |0th| | |
1643 | | entry headers | | | | | | | |
1644 | |_______________|___|_____|________|_______|___| | |
1645 | <---- directory entries ------> | |
1646 | ||
1647 | First directory item has k_offset component 1. We store "." and ".." | |
1648 | in one item, always, we never split "." and ".." into differing | |
1649 | items. This makes, among other things, the code for removing | |
1650 | directories simpler. */ | |
1651 | #define SD_OFFSET 0 | |
1652 | #define SD_UNIQUENESS 0 | |
1653 | #define DOT_OFFSET 1 | |
1654 | #define DOT_DOT_OFFSET 2 | |
1655 | #define DIRENTRY_UNIQUENESS 500 | |
1656 | ||
1657 | /* */ | |
1658 | #define FIRST_ITEM_OFFSET 1 | |
1659 | ||
1660 | /* | |
1661 | Q: How to get key of object pointed to by entry from entry? | |
1662 | ||
1663 | A: Each directory entry has its header. This header has deh_dir_id and deh_objectid fields, those are key | |
1664 | of object, entry points to */ | |
1665 | ||
1666 | /* NOT IMPLEMENTED: | |
1667 | Directory will someday contain stat data of object */ | |
1668 | ||
1669 | struct reiserfs_de_head { | |
1670 | __le32 deh_offset; /* third component of the directory entry key */ | |
1671 | __le32 deh_dir_id; /* objectid of the parent directory of the object, that is referenced | |
1672 | by directory entry */ | |
1673 | __le32 deh_objectid; /* objectid of the object, that is referenced by directory entry */ | |
1674 | __le16 deh_location; /* offset of name in the whole item */ | |
1675 | __le16 deh_state; /* whether 1) entry contains stat data (for future), and 2) whether | |
1676 | entry is hidden (unlinked) */ | |
1677 | } __attribute__ ((__packed__)); | |
1678 | #define DEH_SIZE sizeof(struct reiserfs_de_head) | |
1679 | #define deh_offset(p_deh) (le32_to_cpu((p_deh)->deh_offset)) | |
1680 | #define deh_dir_id(p_deh) (le32_to_cpu((p_deh)->deh_dir_id)) | |
1681 | #define deh_objectid(p_deh) (le32_to_cpu((p_deh)->deh_objectid)) | |
1682 | #define deh_location(p_deh) (le16_to_cpu((p_deh)->deh_location)) | |
1683 | #define deh_state(p_deh) (le16_to_cpu((p_deh)->deh_state)) | |
1684 | ||
1685 | #define put_deh_offset(p_deh,v) ((p_deh)->deh_offset = cpu_to_le32((v))) | |
1686 | #define put_deh_dir_id(p_deh,v) ((p_deh)->deh_dir_id = cpu_to_le32((v))) | |
1687 | #define put_deh_objectid(p_deh,v) ((p_deh)->deh_objectid = cpu_to_le32((v))) | |
1688 | #define put_deh_location(p_deh,v) ((p_deh)->deh_location = cpu_to_le16((v))) | |
1689 | #define put_deh_state(p_deh,v) ((p_deh)->deh_state = cpu_to_le16((v))) | |
1690 | ||
1691 | /* empty directory contains two entries "." and ".." and their headers */ | |
1692 | #define EMPTY_DIR_SIZE \ | |
1693 | (DEH_SIZE * 2 + ROUND_UP (strlen (".")) + ROUND_UP (strlen (".."))) | |
1694 | ||
1695 | /* old format directories have this size when empty */ | |
1696 | #define EMPTY_DIR_SIZE_V1 (DEH_SIZE * 2 + 3) | |
1697 | ||
1698 | #define DEH_Statdata 0 /* not used now */ | |
1699 | #define DEH_Visible 2 | |
1700 | ||
1701 | /* 64 bit systems (and the S/390) need to be aligned explicitly -jdm */ | |
1702 | #if BITS_PER_LONG == 64 || defined(__s390__) || defined(__hppa__) | |
1703 | # define ADDR_UNALIGNED_BITS (3) | |
1704 | #endif | |
1705 | ||
1706 | /* These are only used to manipulate deh_state. | |
1707 | * Because of this, we'll use the ext2_ bit routines, | |
1708 | * since they are little endian */ | |
1709 | #ifdef ADDR_UNALIGNED_BITS | |
1710 | ||
1711 | # define aligned_address(addr) ((void *)((long)(addr) & ~((1UL << ADDR_UNALIGNED_BITS) - 1))) | |
1712 | # define unaligned_offset(addr) (((int)((long)(addr) & ((1 << ADDR_UNALIGNED_BITS) - 1))) << 3) | |
1713 | ||
1714 | # define set_bit_unaligned(nr, addr) \ | |
1715 | __test_and_set_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) | |
1716 | # define clear_bit_unaligned(nr, addr) \ | |
1717 | __test_and_clear_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) | |
1718 | # define test_bit_unaligned(nr, addr) \ | |
1719 | test_bit_le((nr) + unaligned_offset(addr), aligned_address(addr)) | |
1720 | ||
1721 | #else | |
1722 | ||
1723 | # define set_bit_unaligned(nr, addr) __test_and_set_bit_le(nr, addr) | |
1724 | # define clear_bit_unaligned(nr, addr) __test_and_clear_bit_le(nr, addr) | |
1725 | # define test_bit_unaligned(nr, addr) test_bit_le(nr, addr) | |
1726 | ||
1727 | #endif | |
1728 | ||
1729 | #define mark_de_with_sd(deh) set_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) | |
1730 | #define mark_de_without_sd(deh) clear_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) | |
1731 | #define mark_de_visible(deh) set_bit_unaligned (DEH_Visible, &((deh)->deh_state)) | |
1732 | #define mark_de_hidden(deh) clear_bit_unaligned (DEH_Visible, &((deh)->deh_state)) | |
1733 | ||
1734 | #define de_with_sd(deh) test_bit_unaligned (DEH_Statdata, &((deh)->deh_state)) | |
1735 | #define de_visible(deh) test_bit_unaligned (DEH_Visible, &((deh)->deh_state)) | |
1736 | #define de_hidden(deh) !test_bit_unaligned (DEH_Visible, &((deh)->deh_state)) | |
1737 | ||
1738 | extern void make_empty_dir_item_v1(char *body, __le32 dirid, __le32 objid, | |
1739 | __le32 par_dirid, __le32 par_objid); | |
1740 | extern void make_empty_dir_item(char *body, __le32 dirid, __le32 objid, | |
1741 | __le32 par_dirid, __le32 par_objid); | |
1742 | ||
f466c6fd AV |
1743 | // two entries per block (at least) |
1744 | #define REISERFS_MAX_NAME(block_size) 255 | |
1745 | ||
1746 | /* this structure is used for operations on directory entries. It is | |
1747 | not a disk structure. */ | |
1748 | /* When reiserfs_find_entry or search_by_entry_key find directory | |
1749 | entry, they return filled reiserfs_dir_entry structure */ | |
1750 | struct reiserfs_dir_entry { | |
1751 | struct buffer_head *de_bh; | |
1752 | int de_item_num; | |
1753 | struct item_head *de_ih; | |
1754 | int de_entry_num; | |
1755 | struct reiserfs_de_head *de_deh; | |
1756 | int de_entrylen; | |
1757 | int de_namelen; | |
1758 | char *de_name; | |
1759 | unsigned long *de_gen_number_bit_string; | |
1760 | ||
1761 | __u32 de_dir_id; | |
1762 | __u32 de_objectid; | |
1763 | ||
1764 | struct cpu_key de_entry_key; | |
1765 | }; | |
1766 | ||
1767 | /* these defines are useful when a particular member of a reiserfs_dir_entry is needed */ | |
1768 | ||
1769 | /* pointer to file name, stored in entry */ | |
4cf5f7ad JM |
1770 | #define B_I_DEH_ENTRY_FILE_NAME(bh, ih, deh) \ |
1771 | (ih_item_body(bh, ih) + deh_location(deh)) | |
f466c6fd AV |
1772 | |
1773 | /* length of name */ | |
1774 | #define I_DEH_N_ENTRY_FILE_NAME_LENGTH(ih,deh,entry_num) \ | |
1775 | (I_DEH_N_ENTRY_LENGTH (ih, deh, entry_num) - (de_with_sd (deh) ? SD_SIZE : 0)) | |
1776 | ||
1777 | /* hash value occupies bits from 7 up to 30 */ | |
1778 | #define GET_HASH_VALUE(offset) ((offset) & 0x7fffff80LL) | |
1779 | /* generation number occupies 7 bits starting from 0 up to 6 */ | |
1780 | #define GET_GENERATION_NUMBER(offset) ((offset) & 0x7fLL) | |
1781 | #define MAX_GENERATION_NUMBER 127 | |
1782 | ||
1783 | #define SET_GENERATION_NUMBER(offset,gen_number) (GET_HASH_VALUE(offset)|(gen_number)) | |
1784 | ||
1785 | /* | |
1786 | * Picture represents an internal node of the reiserfs tree | |
1787 | * ______________________________________________________ | |
1788 | * | | Array of | Array of | Free | | |
1789 | * |block | keys | pointers | space | | |
1790 | * | head | N | N+1 | | | |
1791 | * |______|_______________|___________________|___________| | |
1792 | */ | |
1793 | ||
1794 | /***************************************************************************/ | |
1795 | /* DISK CHILD */ | |
1796 | /***************************************************************************/ | |
1797 | /* Disk child pointer: The pointer from an internal node of the tree | |
1798 | to a node that is on disk. */ | |
1799 | struct disk_child { | |
1800 | __le32 dc_block_number; /* Disk child's block number. */ | |
1801 | __le16 dc_size; /* Disk child's used space. */ | |
1802 | __le16 dc_reserved; | |
1803 | }; | |
1804 | ||
1805 | #define DC_SIZE (sizeof(struct disk_child)) | |
1806 | #define dc_block_number(dc_p) (le32_to_cpu((dc_p)->dc_block_number)) | |
1807 | #define dc_size(dc_p) (le16_to_cpu((dc_p)->dc_size)) | |
1808 | #define put_dc_block_number(dc_p, val) do { (dc_p)->dc_block_number = cpu_to_le32(val); } while(0) | |
1809 | #define put_dc_size(dc_p, val) do { (dc_p)->dc_size = cpu_to_le16(val); } while(0) | |
1810 | ||
1811 | /* Get disk child by buffer header and position in the tree node. */ | |
1812 | #define B_N_CHILD(bh, n_pos) ((struct disk_child *)\ | |
1813 | ((bh)->b_data + BLKH_SIZE + B_NR_ITEMS(bh) * KEY_SIZE + DC_SIZE * (n_pos))) | |
1814 | ||
1815 | /* Get disk child number by buffer header and position in the tree node. */ | |
1816 | #define B_N_CHILD_NUM(bh, n_pos) (dc_block_number(B_N_CHILD(bh, n_pos))) | |
1817 | #define PUT_B_N_CHILD_NUM(bh, n_pos, val) \ | |
1818 | (put_dc_block_number(B_N_CHILD(bh, n_pos), val)) | |
1819 | ||
1820 | /* maximal value of field child_size in structure disk_child */ | |
1821 | /* child size is the combined size of all items and their headers */ | |
1822 | #define MAX_CHILD_SIZE(bh) ((int)( (bh)->b_size - BLKH_SIZE )) | |
1823 | ||
1824 | /* amount of used space in buffer (not including block head) */ | |
1825 | #define B_CHILD_SIZE(cur) (MAX_CHILD_SIZE(cur)-(B_FREE_SPACE(cur))) | |
1826 | ||
1827 | /* max and min number of keys in internal node */ | |
1828 | #define MAX_NR_KEY(bh) ( (MAX_CHILD_SIZE(bh)-DC_SIZE)/(KEY_SIZE+DC_SIZE) ) | |
1829 | #define MIN_NR_KEY(bh) (MAX_NR_KEY(bh)/2) | |
1830 | ||
1831 | /***************************************************************************/ | |
1832 | /* PATH STRUCTURES AND DEFINES */ | |
1833 | /***************************************************************************/ | |
1834 | ||
1835 | /* Search_by_key fills up the path from the root to the leaf as it descends the tree looking for the | |
1836 | key. It uses reiserfs_bread to try to find buffers in the cache given their block number. If it | |
1837 | does not find them in the cache it reads them from disk. For each node search_by_key finds using | |
1838 | reiserfs_bread it then uses bin_search to look through that node. bin_search will find the | |
1839 | position of the block_number of the next node if it is looking through an internal node. If it | |
1840 | is looking through a leaf node bin_search will find the position of the item which has key either | |
1841 | equal to given key, or which is the maximal key less than the given key. */ | |
1842 | ||
1843 | struct path_element { | |
1844 | struct buffer_head *pe_buffer; /* Pointer to the buffer at the path in the tree. */ | |
1845 | int pe_position; /* Position in the tree node which is placed in the */ | |
1846 | /* buffer above. */ | |
1847 | }; | |
1848 | ||
1849 | #define MAX_HEIGHT 5 /* maximal height of a tree. don't change this without changing JOURNAL_PER_BALANCE_CNT */ | |
1850 | #define EXTENDED_MAX_HEIGHT 7 /* Must be equals MAX_HEIGHT + FIRST_PATH_ELEMENT_OFFSET */ | |
1851 | #define FIRST_PATH_ELEMENT_OFFSET 2 /* Must be equal to at least 2. */ | |
1852 | ||
1853 | #define ILLEGAL_PATH_ELEMENT_OFFSET 1 /* Must be equal to FIRST_PATH_ELEMENT_OFFSET - 1 */ | |
1854 | #define MAX_FEB_SIZE 6 /* this MUST be MAX_HEIGHT + 1. See about FEB below */ | |
1855 | ||
1856 | /* We need to keep track of who the ancestors of nodes are. When we | |
1857 | perform a search we record which nodes were visited while | |
1858 | descending the tree looking for the node we searched for. This list | |
1859 | of nodes is called the path. This information is used while | |
1860 | performing balancing. Note that this path information may become | |
1861 | invalid, and this means we must check it when using it to see if it | |
1862 | is still valid. You'll need to read search_by_key and the comments | |
1863 | in it, especially about decrement_counters_in_path(), to understand | |
1864 | this structure. | |
1865 | ||
1866 | Paths make the code so much harder to work with and debug.... An | |
1867 | enormous number of bugs are due to them, and trying to write or modify | |
1868 | code that uses them just makes my head hurt. They are based on an | |
1869 | excessive effort to avoid disturbing the precious VFS code.:-( The | |
1870 | gods only know how we are going to SMP the code that uses them. | |
1871 | znodes are the way! */ | |
1872 | ||
1873 | #define PATH_READA 0x1 /* do read ahead */ | |
1874 | #define PATH_READA_BACK 0x2 /* read backwards */ | |
1875 | ||
1876 | struct treepath { | |
1877 | int path_length; /* Length of the array above. */ | |
1878 | int reada; | |
1879 | struct path_element path_elements[EXTENDED_MAX_HEIGHT]; /* Array of the path elements. */ | |
1880 | int pos_in_item; | |
1881 | }; | |
1882 | ||
1883 | #define pos_in_item(path) ((path)->pos_in_item) | |
1884 | ||
1885 | #define INITIALIZE_PATH(var) \ | |
1886 | struct treepath var = {.path_length = ILLEGAL_PATH_ELEMENT_OFFSET, .reada = 0,} | |
1887 | ||
1888 | /* Get path element by path and path position. */ | |
1889 | #define PATH_OFFSET_PELEMENT(path, n_offset) ((path)->path_elements + (n_offset)) | |
1890 | ||
1891 | /* Get buffer header at the path by path and path position. */ | |
1892 | #define PATH_OFFSET_PBUFFER(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_buffer) | |
1893 | ||
1894 | /* Get position in the element at the path by path and path position. */ | |
1895 | #define PATH_OFFSET_POSITION(path, n_offset) (PATH_OFFSET_PELEMENT(path, n_offset)->pe_position) | |
1896 | ||
1897 | #define PATH_PLAST_BUFFER(path) (PATH_OFFSET_PBUFFER((path), (path)->path_length)) | |
1898 | /* you know, to the person who didn't | |
1899 | write this the macro name does not | |
1900 | at first suggest what it does. | |
1901 | Maybe POSITION_FROM_PATH_END? Or | |
1902 | maybe we should just focus on | |
1903 | dumping paths... -Hans */ | |
1904 | #define PATH_LAST_POSITION(path) (PATH_OFFSET_POSITION((path), (path)->path_length)) | |
1905 | ||
f466c6fd AV |
1906 | /* in do_balance leaf has h == 0 in contrast with path structure, |
1907 | where root has level == 0. That is why we need these defines */ | |
1908 | #define PATH_H_PBUFFER(path, h) PATH_OFFSET_PBUFFER (path, path->path_length - (h)) /* tb->S[h] */ | |
1909 | #define PATH_H_PPARENT(path, h) PATH_H_PBUFFER (path, (h) + 1) /* tb->F[h] or tb->S[0]->b_parent */ | |
1910 | #define PATH_H_POSITION(path, h) PATH_OFFSET_POSITION (path, path->path_length - (h)) | |
1911 | #define PATH_H_B_ITEM_ORDER(path, h) PATH_H_POSITION(path, h + 1) /* tb->S[h]->b_item_order */ | |
1912 | ||
1913 | #define PATH_H_PATH_OFFSET(path, n_h) ((path)->path_length - (n_h)) | |
1914 | ||
4cf5f7ad JM |
1915 | static inline void *reiserfs_node_data(const struct buffer_head *bh) |
1916 | { | |
1917 | return bh->b_data + sizeof(struct block_head); | |
1918 | } | |
1919 | ||
1920 | /* get key from internal node */ | |
1921 | static inline struct reiserfs_key *internal_key(struct buffer_head *bh, | |
1922 | int item_num) | |
1923 | { | |
1924 | struct reiserfs_key *key = reiserfs_node_data(bh); | |
1925 | ||
1926 | return &key[item_num]; | |
1927 | } | |
1928 | ||
1929 | /* get the item header from leaf node */ | |
1930 | static inline struct item_head *item_head(const struct buffer_head *bh, | |
1931 | int item_num) | |
1932 | { | |
1933 | struct item_head *ih = reiserfs_node_data(bh); | |
1934 | ||
1935 | return &ih[item_num]; | |
1936 | } | |
1937 | ||
1938 | /* get the key from leaf node */ | |
1939 | static inline struct reiserfs_key *leaf_key(const struct buffer_head *bh, | |
1940 | int item_num) | |
1941 | { | |
1942 | return &item_head(bh, item_num)->ih_key; | |
1943 | } | |
1944 | ||
1945 | static inline void *ih_item_body(const struct buffer_head *bh, | |
1946 | const struct item_head *ih) | |
1947 | { | |
1948 | return bh->b_data + ih_location(ih); | |
1949 | } | |
1950 | ||
1951 | /* get item body from leaf node */ | |
1952 | static inline void *item_body(const struct buffer_head *bh, int item_num) | |
1953 | { | |
1954 | return ih_item_body(bh, item_head(bh, item_num)); | |
1955 | } | |
1956 | ||
1957 | static inline struct item_head *tp_item_head(const struct treepath *path) | |
1958 | { | |
1959 | return item_head(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path)); | |
1960 | } | |
1961 | ||
1962 | static inline void *tp_item_body(const struct treepath *path) | |
1963 | { | |
1964 | return item_body(PATH_PLAST_BUFFER(path), PATH_LAST_POSITION(path)); | |
1965 | } | |
1966 | ||
f466c6fd | 1967 | #define get_last_bh(path) PATH_PLAST_BUFFER(path) |
f466c6fd | 1968 | #define get_item_pos(path) PATH_LAST_POSITION(path) |
f466c6fd AV |
1969 | #define item_moved(ih,path) comp_items(ih, path) |
1970 | #define path_changed(ih,path) comp_items (ih, path) | |
1971 | ||
4cf5f7ad JM |
1972 | /* array of the entry headers */ |
1973 | /* get item body */ | |
1974 | #define B_I_DEH(bh, ih) ((struct reiserfs_de_head *)(ih_item_body(bh, ih))) | |
1975 | ||
1976 | /* length of the directory entry in directory item. This define | |
1977 | calculates length of i-th directory entry using directory entry | |
1978 | locations from dir entry head. When it calculates length of 0-th | |
1979 | directory entry, it uses length of whole item in place of entry | |
1980 | location of the non-existent following entry in the calculation. | |
1981 | See picture above.*/ | |
1982 | /* | |
1983 | #define I_DEH_N_ENTRY_LENGTH(ih,deh,i) \ | |
1984 | ((i) ? (deh_location((deh)-1) - deh_location((deh))) : (ih_item_len((ih)) - deh_location((deh)))) | |
1985 | */ | |
1986 | static inline int entry_length(const struct buffer_head *bh, | |
1987 | const struct item_head *ih, int pos_in_item) | |
1988 | { | |
1989 | struct reiserfs_de_head *deh; | |
1990 | ||
1991 | deh = B_I_DEH(bh, ih) + pos_in_item; | |
1992 | if (pos_in_item) | |
1993 | return deh_location(deh - 1) - deh_location(deh); | |
1994 | ||
1995 | return ih_item_len(ih) - deh_location(deh); | |
1996 | } | |
1997 | ||
f466c6fd AV |
1998 | /***************************************************************************/ |
1999 | /* MISC */ | |
2000 | /***************************************************************************/ | |
2001 | ||
2002 | /* Size of pointer to the unformatted node. */ | |
2003 | #define UNFM_P_SIZE (sizeof(unp_t)) | |
2004 | #define UNFM_P_SHIFT 2 | |
2005 | ||
2006 | // in in-core inode key is stored on le form | |
2007 | #define INODE_PKEY(inode) ((struct reiserfs_key *)(REISERFS_I(inode)->i_key)) | |
2008 | ||
2009 | #define MAX_UL_INT 0xffffffff | |
2010 | #define MAX_INT 0x7ffffff | |
2011 | #define MAX_US_INT 0xffff | |
2012 | ||
2013 | // reiserfs version 2 has max offset 60 bits. Version 1 - 32 bit offset | |
f466c6fd AV |
2014 | static inline loff_t max_reiserfs_offset(struct inode *inode) |
2015 | { | |
2016 | if (get_inode_item_key_version(inode) == KEY_FORMAT_3_5) | |
2017 | return (loff_t) U32_MAX; | |
2018 | ||
2019 | return (loff_t) ((~(__u64) 0) >> 4); | |
2020 | } | |
2021 | ||
2022 | /*#define MAX_KEY_UNIQUENESS MAX_UL_INT*/ | |
2023 | #define MAX_KEY_OBJECTID MAX_UL_INT | |
2024 | ||
2025 | #define MAX_B_NUM MAX_UL_INT | |
2026 | #define MAX_FC_NUM MAX_US_INT | |
2027 | ||
2028 | /* the purpose is to detect overflow of an unsigned short */ | |
2029 | #define REISERFS_LINK_MAX (MAX_US_INT - 1000) | |
2030 | ||
2031 | /* The following defines are used in reiserfs_insert_item and reiserfs_append_item */ | |
2032 | #define REISERFS_KERNEL_MEM 0 /* reiserfs kernel memory mode */ | |
2033 | #define REISERFS_USER_MEM 1 /* reiserfs user memory mode */ | |
2034 | ||
2035 | #define fs_generation(s) (REISERFS_SB(s)->s_generation_counter) | |
2036 | #define get_generation(s) atomic_read (&fs_generation(s)) | |
2037 | #define FILESYSTEM_CHANGED_TB(tb) (get_generation((tb)->tb_sb) != (tb)->fs_gen) | |
2038 | #define __fs_changed(gen,s) (gen != get_generation (s)) | |
2039 | #define fs_changed(gen,s) \ | |
2040 | ({ \ | |
2041 | reiserfs_cond_resched(s); \ | |
2042 | __fs_changed(gen, s); \ | |
2043 | }) | |
2044 | ||
2045 | /***************************************************************************/ | |
2046 | /* FIXATE NODES */ | |
2047 | /***************************************************************************/ | |
2048 | ||
2049 | #define VI_TYPE_LEFT_MERGEABLE 1 | |
2050 | #define VI_TYPE_RIGHT_MERGEABLE 2 | |
2051 | ||
2052 | /* To make any changes in the tree we always first find node, that | |
2053 | contains item to be changed/deleted or place to insert a new | |
2054 | item. We call this node S. To do balancing we need to decide what | |
2055 | we will shift to left/right neighbor, or to a new node, where new | |
2056 | item will be etc. To make this analysis simpler we build virtual | |
2057 | node. Virtual node is an array of items, that will replace items of | |
2058 | node S. (For instance if we are going to delete an item, virtual | |
2059 | node does not contain it). Virtual node keeps information about | |
2060 | item sizes and types, mergeability of first and last items, sizes | |
2061 | of all entries in directory item. We use this array of items when | |
2062 | calculating what we can shift to neighbors and how many nodes we | |
2063 | have to have if we do not any shiftings, if we shift to left/right | |
2064 | neighbor or to both. */ | |
2065 | struct virtual_item { | |
2066 | int vi_index; // index in the array of item operations | |
2067 | unsigned short vi_type; // left/right mergeability | |
2068 | unsigned short vi_item_len; /* length of item that it will have after balancing */ | |
2069 | struct item_head *vi_ih; | |
2070 | const char *vi_item; // body of item (old or new) | |
2071 | const void *vi_new_data; // 0 always but paste mode | |
2072 | void *vi_uarea; // item specific area | |
2073 | }; | |
2074 | ||
2075 | struct virtual_node { | |
2076 | char *vn_free_ptr; /* this is a pointer to the free space in the buffer */ | |
2077 | unsigned short vn_nr_item; /* number of items in virtual node */ | |
2078 | short vn_size; /* size of node , that node would have if it has unlimited size and no balancing is performed */ | |
2079 | short vn_mode; /* mode of balancing (paste, insert, delete, cut) */ | |
2080 | short vn_affected_item_num; | |
2081 | short vn_pos_in_item; | |
2082 | struct item_head *vn_ins_ih; /* item header of inserted item, 0 for other modes */ | |
2083 | const void *vn_data; | |
2084 | struct virtual_item *vn_vi; /* array of items (including a new one, excluding item to be deleted) */ | |
2085 | }; | |
2086 | ||
2087 | /* used by directory items when creating virtual nodes */ | |
2088 | struct direntry_uarea { | |
2089 | int flags; | |
2090 | __u16 entry_count; | |
2091 | __u16 entry_sizes[1]; | |
2092 | } __attribute__ ((__packed__)); | |
2093 | ||
2094 | /***************************************************************************/ | |
2095 | /* TREE BALANCE */ | |
2096 | /***************************************************************************/ | |
2097 | ||
2098 | /* This temporary structure is used in tree balance algorithms, and | |
2099 | constructed as we go to the extent that its various parts are | |
2100 | needed. It contains arrays of nodes that can potentially be | |
2101 | involved in the balancing of node S, and parameters that define how | |
2102 | each of the nodes must be balanced. Note that in these algorithms | |
2103 | for balancing the worst case is to need to balance the current node | |
2104 | S and the left and right neighbors and all of their parents plus | |
2105 | create a new node. We implement S1 balancing for the leaf nodes | |
2106 | and S0 balancing for the internal nodes (S1 and S0 are defined in | |
2107 | our papers.)*/ | |
2108 | ||
2109 | #define MAX_FREE_BLOCK 7 /* size of the array of buffers to free at end of do_balance */ | |
2110 | ||
2111 | /* maximum number of FEB blocknrs on a single level */ | |
2112 | #define MAX_AMOUNT_NEEDED 2 | |
2113 | ||
2114 | /* someday somebody will prefix every field in this struct with tb_ */ | |
2115 | struct tree_balance { | |
2116 | int tb_mode; | |
2117 | int need_balance_dirty; | |
2118 | struct super_block *tb_sb; | |
2119 | struct reiserfs_transaction_handle *transaction_handle; | |
2120 | struct treepath *tb_path; | |
2121 | struct buffer_head *L[MAX_HEIGHT]; /* array of left neighbors of nodes in the path */ | |
2122 | struct buffer_head *R[MAX_HEIGHT]; /* array of right neighbors of nodes in the path */ | |
2123 | struct buffer_head *FL[MAX_HEIGHT]; /* array of fathers of the left neighbors */ | |
2124 | struct buffer_head *FR[MAX_HEIGHT]; /* array of fathers of the right neighbors */ | |
2125 | struct buffer_head *CFL[MAX_HEIGHT]; /* array of common parents of center node and its left neighbor */ | |
2126 | struct buffer_head *CFR[MAX_HEIGHT]; /* array of common parents of center node and its right neighbor */ | |
2127 | ||
2128 | struct buffer_head *FEB[MAX_FEB_SIZE]; /* array of empty buffers. Number of buffers in array equals | |
2129 | cur_blknum. */ | |
2130 | struct buffer_head *used[MAX_FEB_SIZE]; | |
2131 | struct buffer_head *thrown[MAX_FEB_SIZE]; | |
2132 | int lnum[MAX_HEIGHT]; /* array of number of items which must be | |
2133 | shifted to the left in order to balance the | |
2134 | current node; for leaves includes item that | |
2135 | will be partially shifted; for internal | |
2136 | nodes, it is the number of child pointers | |
2137 | rather than items. It includes the new item | |
2138 | being created. The code sometimes subtracts | |
2139 | one to get the number of wholly shifted | |
2140 | items for other purposes. */ | |
2141 | int rnum[MAX_HEIGHT]; /* substitute right for left in comment above */ | |
2142 | int lkey[MAX_HEIGHT]; /* array indexed by height h mapping the key delimiting L[h] and | |
2143 | S[h] to its item number within the node CFL[h] */ | |
2144 | int rkey[MAX_HEIGHT]; /* substitute r for l in comment above */ | |
2145 | int insert_size[MAX_HEIGHT]; /* the number of bytes by we are trying to add or remove from | |
2146 | S[h]. A negative value means removing. */ | |
2147 | int blknum[MAX_HEIGHT]; /* number of nodes that will replace node S[h] after | |
2148 | balancing on the level h of the tree. If 0 then S is | |
2149 | being deleted, if 1 then S is remaining and no new nodes | |
2150 | are being created, if 2 or 3 then 1 or 2 new nodes is | |
2151 | being created */ | |
2152 | ||
2153 | /* fields that are used only for balancing leaves of the tree */ | |
2154 | int cur_blknum; /* number of empty blocks having been already allocated */ | |
2155 | int s0num; /* number of items that fall into left most node when S[0] splits */ | |
2156 | int s1num; /* number of items that fall into first new node when S[0] splits */ | |
2157 | int s2num; /* number of items that fall into second new node when S[0] splits */ | |
2158 | int lbytes; /* number of bytes which can flow to the left neighbor from the left */ | |
2159 | /* most liquid item that cannot be shifted from S[0] entirely */ | |
2160 | /* if -1 then nothing will be partially shifted */ | |
2161 | int rbytes; /* number of bytes which will flow to the right neighbor from the right */ | |
2162 | /* most liquid item that cannot be shifted from S[0] entirely */ | |
2163 | /* if -1 then nothing will be partially shifted */ | |
2164 | int s1bytes; /* number of bytes which flow to the first new node when S[0] splits */ | |
2165 | /* note: if S[0] splits into 3 nodes, then items do not need to be cut */ | |
2166 | int s2bytes; | |
2167 | struct buffer_head *buf_to_free[MAX_FREE_BLOCK]; /* buffers which are to be freed after do_balance finishes by unfix_nodes */ | |
2168 | char *vn_buf; /* kmalloced memory. Used to create | |
2169 | virtual node and keep map of | |
2170 | dirtied bitmap blocks */ | |
2171 | int vn_buf_size; /* size of the vn_buf */ | |
2172 | struct virtual_node *tb_vn; /* VN starts after bitmap of bitmap blocks */ | |
2173 | ||
2174 | int fs_gen; /* saved value of `reiserfs_generation' counter | |
2175 | see FILESYSTEM_CHANGED() macro in reiserfs_fs.h */ | |
2176 | #ifdef DISPLACE_NEW_PACKING_LOCALITIES | |
2177 | struct in_core_key key; /* key pointer, to pass to block allocator or | |
2178 | another low-level subsystem */ | |
2179 | #endif | |
2180 | }; | |
2181 | ||
2182 | /* These are modes of balancing */ | |
2183 | ||
2184 | /* When inserting an item. */ | |
2185 | #define M_INSERT 'i' | |
2186 | /* When inserting into (directories only) or appending onto an already | |
2187 | existent item. */ | |
2188 | #define M_PASTE 'p' | |
2189 | /* When deleting an item. */ | |
2190 | #define M_DELETE 'd' | |
2191 | /* When truncating an item or removing an entry from a (directory) item. */ | |
2192 | #define M_CUT 'c' | |
2193 | ||
2194 | /* used when balancing on leaf level skipped (in reiserfsck) */ | |
2195 | #define M_INTERNAL 'n' | |
2196 | ||
2197 | /* When further balancing is not needed, then do_balance does not need | |
2198 | to be called. */ | |
2199 | #define M_SKIP_BALANCING 's' | |
2200 | #define M_CONVERT 'v' | |
2201 | ||
2202 | /* modes of leaf_move_items */ | |
2203 | #define LEAF_FROM_S_TO_L 0 | |
2204 | #define LEAF_FROM_S_TO_R 1 | |
2205 | #define LEAF_FROM_R_TO_L 2 | |
2206 | #define LEAF_FROM_L_TO_R 3 | |
2207 | #define LEAF_FROM_S_TO_SNEW 4 | |
2208 | ||
2209 | #define FIRST_TO_LAST 0 | |
2210 | #define LAST_TO_FIRST 1 | |
2211 | ||
2212 | /* used in do_balance for passing parent of node information that has | |
2213 | been gotten from tb struct */ | |
2214 | struct buffer_info { | |
2215 | struct tree_balance *tb; | |
2216 | struct buffer_head *bi_bh; | |
2217 | struct buffer_head *bi_parent; | |
2218 | int bi_position; | |
2219 | }; | |
2220 | ||
2221 | static inline struct super_block *sb_from_tb(struct tree_balance *tb) | |
2222 | { | |
2223 | return tb ? tb->tb_sb : NULL; | |
2224 | } | |
2225 | ||
2226 | static inline struct super_block *sb_from_bi(struct buffer_info *bi) | |
2227 | { | |
2228 | return bi ? sb_from_tb(bi->tb) : NULL; | |
2229 | } | |
2230 | ||
2231 | /* there are 4 types of items: stat data, directory item, indirect, direct. | |
2232 | +-------------------+------------+--------------+------------+ | |
2233 | | | k_offset | k_uniqueness | mergeable? | | |
2234 | +-------------------+------------+--------------+------------+ | |
2235 | | stat data | 0 | 0 | no | | |
2236 | +-------------------+------------+--------------+------------+ | |
2237 | | 1st directory item| DOT_OFFSET |DIRENTRY_UNIQUENESS| no | | |
2238 | | non 1st directory | hash value | | yes | | |
2239 | | item | | | | | |
2240 | +-------------------+------------+--------------+------------+ | |
2241 | | indirect item | offset + 1 |TYPE_INDIRECT | if this is not the first indirect item of the object | |
2242 | +-------------------+------------+--------------+------------+ | |
2243 | | direct item | offset + 1 |TYPE_DIRECT | if not this is not the first direct item of the object | |
2244 | +-------------------+------------+--------------+------------+ | |
2245 | */ | |
2246 | ||
2247 | struct item_operations { | |
2248 | int (*bytes_number) (struct item_head * ih, int block_size); | |
2249 | void (*decrement_key) (struct cpu_key *); | |
2250 | int (*is_left_mergeable) (struct reiserfs_key * ih, | |
2251 | unsigned long bsize); | |
2252 | void (*print_item) (struct item_head *, char *item); | |
2253 | void (*check_item) (struct item_head *, char *item); | |
2254 | ||
2255 | int (*create_vi) (struct virtual_node * vn, struct virtual_item * vi, | |
2256 | int is_affected, int insert_size); | |
2257 | int (*check_left) (struct virtual_item * vi, int free, | |
2258 | int start_skip, int end_skip); | |
2259 | int (*check_right) (struct virtual_item * vi, int free); | |
2260 | int (*part_size) (struct virtual_item * vi, int from, int to); | |
2261 | int (*unit_num) (struct virtual_item * vi); | |
2262 | void (*print_vi) (struct virtual_item * vi); | |
2263 | }; | |
2264 | ||
2265 | extern struct item_operations *item_ops[TYPE_ANY + 1]; | |
2266 | ||
2267 | #define op_bytes_number(ih,bsize) item_ops[le_ih_k_type (ih)]->bytes_number (ih, bsize) | |
2268 | #define op_is_left_mergeable(key,bsize) item_ops[le_key_k_type (le_key_version (key), key)]->is_left_mergeable (key, bsize) | |
2269 | #define op_print_item(ih,item) item_ops[le_ih_k_type (ih)]->print_item (ih, item) | |
2270 | #define op_check_item(ih,item) item_ops[le_ih_k_type (ih)]->check_item (ih, item) | |
2271 | #define op_create_vi(vn,vi,is_affected,insert_size) item_ops[le_ih_k_type ((vi)->vi_ih)]->create_vi (vn,vi,is_affected,insert_size) | |
2272 | #define op_check_left(vi,free,start_skip,end_skip) item_ops[(vi)->vi_index]->check_left (vi, free, start_skip, end_skip) | |
2273 | #define op_check_right(vi,free) item_ops[(vi)->vi_index]->check_right (vi, free) | |
2274 | #define op_part_size(vi,from,to) item_ops[(vi)->vi_index]->part_size (vi, from, to) | |
2275 | #define op_unit_num(vi) item_ops[(vi)->vi_index]->unit_num (vi) | |
2276 | #define op_print_vi(vi) item_ops[(vi)->vi_index]->print_vi (vi) | |
2277 | ||
2278 | #define COMP_SHORT_KEYS comp_short_keys | |
2279 | ||
2280 | /* number of blocks pointed to by the indirect item */ | |
2281 | #define I_UNFM_NUM(ih) (ih_item_len(ih) / UNFM_P_SIZE) | |
2282 | ||
2283 | /* the used space within the unformatted node corresponding to pos within the item pointed to by ih */ | |
2284 | #define I_POS_UNFM_SIZE(ih,pos,size) (((pos) == I_UNFM_NUM(ih) - 1 ) ? (size) - ih_free_space(ih) : (size)) | |
2285 | ||
2286 | /* number of bytes contained by the direct item or the unformatted nodes the indirect item points to */ | |
2287 | ||
f466c6fd AV |
2288 | /* following defines use reiserfs buffer header and item header */ |
2289 | ||
2290 | /* get stat-data */ | |
2291 | #define B_I_STAT_DATA(bh, ih) ( (struct stat_data * )((bh)->b_data + ih_location(ih)) ) | |
2292 | ||
2293 | // this is 3976 for size==4096 | |
2294 | #define MAX_DIRECT_ITEM_LEN(size) ((size) - BLKH_SIZE - 2*IH_SIZE - SD_SIZE - UNFM_P_SIZE) | |
2295 | ||
2296 | /* indirect items consist of entries which contain blocknrs, pos | |
2297 | indicates which entry, and B_I_POS_UNFM_POINTER resolves to the | |
2298 | blocknr contained by the entry pos points to */ | |
4cf5f7ad JM |
2299 | #define B_I_POS_UNFM_POINTER(bh, ih, pos) \ |
2300 | le32_to_cpu(*(((unp_t *)ih_item_body(bh, ih)) + (pos))) | |
2301 | #define PUT_B_I_POS_UNFM_POINTER(bh, ih, pos, val) \ | |
2302 | (*(((unp_t *)ih_item_body(bh, ih)) + (pos)) = cpu_to_le32(val)) | |
f466c6fd AV |
2303 | |
2304 | struct reiserfs_iget_args { | |
2305 | __u32 objectid; | |
2306 | __u32 dirid; | |
2307 | }; | |
2308 | ||
2309 | /***************************************************************************/ | |
2310 | /* FUNCTION DECLARATIONS */ | |
2311 | /***************************************************************************/ | |
2312 | ||
2313 | #define get_journal_desc_magic(bh) (bh->b_data + bh->b_size - 12) | |
2314 | ||
2315 | #define journal_trans_half(blocksize) \ | |
2316 | ((blocksize - sizeof (struct reiserfs_journal_desc) + sizeof (__u32) - 12) / sizeof (__u32)) | |
2317 | ||
2318 | /* journal.c see journal.c for all the comments here */ | |
2319 | ||
2320 | /* first block written in a commit. */ | |
2321 | struct reiserfs_journal_desc { | |
2322 | __le32 j_trans_id; /* id of commit */ | |
2323 | __le32 j_len; /* length of commit. len +1 is the commit block */ | |
2324 | __le32 j_mount_id; /* mount id of this trans */ | |
2325 | __le32 j_realblock[1]; /* real locations for each block */ | |
2326 | }; | |
2327 | ||
2328 | #define get_desc_trans_id(d) le32_to_cpu((d)->j_trans_id) | |
2329 | #define get_desc_trans_len(d) le32_to_cpu((d)->j_len) | |
2330 | #define get_desc_mount_id(d) le32_to_cpu((d)->j_mount_id) | |
2331 | ||
2332 | #define set_desc_trans_id(d,val) do { (d)->j_trans_id = cpu_to_le32 (val); } while (0) | |
2333 | #define set_desc_trans_len(d,val) do { (d)->j_len = cpu_to_le32 (val); } while (0) | |
2334 | #define set_desc_mount_id(d,val) do { (d)->j_mount_id = cpu_to_le32 (val); } while (0) | |
2335 | ||
2336 | /* last block written in a commit */ | |
2337 | struct reiserfs_journal_commit { | |
2338 | __le32 j_trans_id; /* must match j_trans_id from the desc block */ | |
2339 | __le32 j_len; /* ditto */ | |
2340 | __le32 j_realblock[1]; /* real locations for each block */ | |
2341 | }; | |
2342 | ||
2343 | #define get_commit_trans_id(c) le32_to_cpu((c)->j_trans_id) | |
2344 | #define get_commit_trans_len(c) le32_to_cpu((c)->j_len) | |
2345 | #define get_commit_mount_id(c) le32_to_cpu((c)->j_mount_id) | |
2346 | ||
2347 | #define set_commit_trans_id(c,val) do { (c)->j_trans_id = cpu_to_le32 (val); } while (0) | |
2348 | #define set_commit_trans_len(c,val) do { (c)->j_len = cpu_to_le32 (val); } while (0) | |
2349 | ||
2350 | /* this header block gets written whenever a transaction is considered fully flushed, and is more recent than the | |
2351 | ** last fully flushed transaction. fully flushed means all the log blocks and all the real blocks are on disk, | |
2352 | ** and this transaction does not need to be replayed. | |
2353 | */ | |
2354 | struct reiserfs_journal_header { | |
2355 | __le32 j_last_flush_trans_id; /* id of last fully flushed transaction */ | |
2356 | __le32 j_first_unflushed_offset; /* offset in the log of where to start replay after a crash */ | |
2357 | __le32 j_mount_id; | |
2358 | /* 12 */ struct journal_params jh_journal; | |
2359 | }; | |
2360 | ||
2361 | /* biggest tunable defines are right here */ | |
2362 | #define JOURNAL_BLOCK_COUNT 8192 /* number of blocks in the journal */ | |
2363 | #define JOURNAL_TRANS_MAX_DEFAULT 1024 /* biggest possible single transaction, don't change for now (8/3/99) */ | |
2364 | #define JOURNAL_TRANS_MIN_DEFAULT 256 | |
2365 | #define JOURNAL_MAX_BATCH_DEFAULT 900 /* max blocks to batch into one transaction, don't make this any bigger than 900 */ | |
2366 | #define JOURNAL_MIN_RATIO 2 | |
2367 | #define JOURNAL_MAX_COMMIT_AGE 30 | |
2368 | #define JOURNAL_MAX_TRANS_AGE 30 | |
2369 | #define JOURNAL_PER_BALANCE_CNT (3 * (MAX_HEIGHT-2) + 9) | |
2370 | #define JOURNAL_BLOCKS_PER_OBJECT(sb) (JOURNAL_PER_BALANCE_CNT * 3 + \ | |
2371 | 2 * (REISERFS_QUOTA_INIT_BLOCKS(sb) + \ | |
2372 | REISERFS_QUOTA_TRANS_BLOCKS(sb))) | |
2373 | ||
2374 | #ifdef CONFIG_QUOTA | |
2375 | #define REISERFS_QUOTA_OPTS ((1 << REISERFS_USRQUOTA) | (1 << REISERFS_GRPQUOTA)) | |
2376 | /* We need to update data and inode (atime) */ | |
2377 | #define REISERFS_QUOTA_TRANS_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? 2 : 0) | |
2378 | /* 1 balancing, 1 bitmap, 1 data per write + stat data update */ | |
2379 | #define REISERFS_QUOTA_INIT_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \ | |
2380 | (DQUOT_INIT_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_INIT_REWRITE+1) : 0) | |
2381 | /* same as with INIT */ | |
2382 | #define REISERFS_QUOTA_DEL_BLOCKS(s) (REISERFS_SB(s)->s_mount_opt & REISERFS_QUOTA_OPTS ? \ | |
2383 | (DQUOT_DEL_ALLOC*(JOURNAL_PER_BALANCE_CNT+2)+DQUOT_DEL_REWRITE+1) : 0) | |
2384 | #else | |
2385 | #define REISERFS_QUOTA_TRANS_BLOCKS(s) 0 | |
2386 | #define REISERFS_QUOTA_INIT_BLOCKS(s) 0 | |
2387 | #define REISERFS_QUOTA_DEL_BLOCKS(s) 0 | |
2388 | #endif | |
2389 | ||
2390 | /* both of these can be as low as 1, or as high as you want. The min is the | |
2391 | ** number of 4k bitmap nodes preallocated on mount. New nodes are allocated | |
2392 | ** as needed, and released when transactions are committed. On release, if | |
2393 | ** the current number of nodes is > max, the node is freed, otherwise, | |
2394 | ** it is put on a free list for faster use later. | |
2395 | */ | |
2396 | #define REISERFS_MIN_BITMAP_NODES 10 | |
2397 | #define REISERFS_MAX_BITMAP_NODES 100 | |
2398 | ||
2399 | #define JBH_HASH_SHIFT 13 /* these are based on journal hash size of 8192 */ | |
2400 | #define JBH_HASH_MASK 8191 | |
2401 | ||
2402 | #define _jhashfn(sb,block) \ | |
2403 | (((unsigned long)sb>>L1_CACHE_SHIFT) ^ \ | |
2404 | (((block)<<(JBH_HASH_SHIFT - 6)) ^ ((block) >> 13) ^ ((block) << (JBH_HASH_SHIFT - 12)))) | |
2405 | #define journal_hash(t,sb,block) ((t)[_jhashfn((sb),(block)) & JBH_HASH_MASK]) | |
2406 | ||
2407 | // We need these to make journal.c code more readable | |
2408 | #define journal_find_get_block(s, block) __find_get_block(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) | |
2409 | #define journal_getblk(s, block) __getblk(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) | |
2410 | #define journal_bread(s, block) __bread(SB_JOURNAL(s)->j_dev_bd, block, s->s_blocksize) | |
2411 | ||
2412 | enum reiserfs_bh_state_bits { | |
2413 | BH_JDirty = BH_PrivateStart, /* buffer is in current transaction */ | |
2414 | BH_JDirty_wait, | |
2415 | BH_JNew, /* disk block was taken off free list before | |
2416 | * being in a finished transaction, or | |
2417 | * written to disk. Can be reused immed. */ | |
2418 | BH_JPrepared, | |
2419 | BH_JRestore_dirty, | |
2420 | BH_JTest, // debugging only will go away | |
2421 | }; | |
2422 | ||
2423 | BUFFER_FNS(JDirty, journaled); | |
2424 | TAS_BUFFER_FNS(JDirty, journaled); | |
2425 | BUFFER_FNS(JDirty_wait, journal_dirty); | |
2426 | TAS_BUFFER_FNS(JDirty_wait, journal_dirty); | |
2427 | BUFFER_FNS(JNew, journal_new); | |
2428 | TAS_BUFFER_FNS(JNew, journal_new); | |
2429 | BUFFER_FNS(JPrepared, journal_prepared); | |
2430 | TAS_BUFFER_FNS(JPrepared, journal_prepared); | |
2431 | BUFFER_FNS(JRestore_dirty, journal_restore_dirty); | |
2432 | TAS_BUFFER_FNS(JRestore_dirty, journal_restore_dirty); | |
2433 | BUFFER_FNS(JTest, journal_test); | |
2434 | TAS_BUFFER_FNS(JTest, journal_test); | |
2435 | ||
2436 | /* | |
2437 | ** transaction handle which is passed around for all journal calls | |
2438 | */ | |
2439 | struct reiserfs_transaction_handle { | |
2440 | struct super_block *t_super; /* super for this FS when journal_begin was | |
2441 | called. saves calls to reiserfs_get_super | |
2442 | also used by nested transactions to make | |
2443 | sure they are nesting on the right FS | |
2444 | _must_ be first in the handle | |
2445 | */ | |
2446 | int t_refcount; | |
2447 | int t_blocks_logged; /* number of blocks this writer has logged */ | |
2448 | int t_blocks_allocated; /* number of blocks this writer allocated */ | |
2449 | unsigned int t_trans_id; /* sanity check, equals the current trans id */ | |
2450 | void *t_handle_save; /* save existing current->journal_info */ | |
2451 | unsigned displace_new_blocks:1; /* if new block allocation occurres, that block | |
2452 | should be displaced from others */ | |
2453 | struct list_head t_list; | |
2454 | }; | |
2455 | ||
2456 | /* used to keep track of ordered and tail writes, attached to the buffer | |
2457 | * head through b_journal_head. | |
2458 | */ | |
2459 | struct reiserfs_jh { | |
2460 | struct reiserfs_journal_list *jl; | |
2461 | struct buffer_head *bh; | |
2462 | struct list_head list; | |
2463 | }; | |
2464 | ||
2465 | void reiserfs_free_jh(struct buffer_head *bh); | |
2466 | int reiserfs_add_tail_list(struct inode *inode, struct buffer_head *bh); | |
2467 | int reiserfs_add_ordered_list(struct inode *inode, struct buffer_head *bh); | |
2468 | int journal_mark_dirty(struct reiserfs_transaction_handle *, | |
2469 | struct super_block *, struct buffer_head *bh); | |
2470 | ||
2471 | static inline int reiserfs_file_data_log(struct inode *inode) | |
2472 | { | |
2473 | if (reiserfs_data_log(inode->i_sb) || | |
2474 | (REISERFS_I(inode)->i_flags & i_data_log)) | |
2475 | return 1; | |
2476 | return 0; | |
2477 | } | |
2478 | ||
2479 | static inline int reiserfs_transaction_running(struct super_block *s) | |
2480 | { | |
2481 | struct reiserfs_transaction_handle *th = current->journal_info; | |
2482 | if (th && th->t_super == s) | |
2483 | return 1; | |
2484 | if (th && th->t_super == NULL) | |
2485 | BUG(); | |
2486 | return 0; | |
2487 | } | |
2488 | ||
2489 | static inline int reiserfs_transaction_free_space(struct reiserfs_transaction_handle *th) | |
2490 | { | |
2491 | return th->t_blocks_allocated - th->t_blocks_logged; | |
2492 | } | |
2493 | ||
2494 | struct reiserfs_transaction_handle *reiserfs_persistent_transaction(struct | |
2495 | super_block | |
2496 | *, | |
2497 | int count); | |
2498 | int reiserfs_end_persistent_transaction(struct reiserfs_transaction_handle *); | |
cfac4b47 | 2499 | void reiserfs_vfs_truncate_file(struct inode *inode); |
f466c6fd AV |
2500 | int reiserfs_commit_page(struct inode *inode, struct page *page, |
2501 | unsigned from, unsigned to); | |
25729b0e | 2502 | void reiserfs_flush_old_commits(struct super_block *); |
f466c6fd AV |
2503 | int reiserfs_commit_for_inode(struct inode *); |
2504 | int reiserfs_inode_needs_commit(struct inode *); | |
2505 | void reiserfs_update_inode_transaction(struct inode *); | |
2506 | void reiserfs_wait_on_write_block(struct super_block *s); | |
2507 | void reiserfs_block_writes(struct reiserfs_transaction_handle *th); | |
2508 | void reiserfs_allow_writes(struct super_block *s); | |
2509 | void reiserfs_check_lock_depth(struct super_block *s, char *caller); | |
2510 | int reiserfs_prepare_for_journal(struct super_block *, struct buffer_head *bh, | |
2511 | int wait); | |
2512 | void reiserfs_restore_prepared_buffer(struct super_block *, | |
2513 | struct buffer_head *bh); | |
2514 | int journal_init(struct super_block *, const char *j_dev_name, int old_format, | |
2515 | unsigned int); | |
2516 | int journal_release(struct reiserfs_transaction_handle *, struct super_block *); | |
2517 | int journal_release_error(struct reiserfs_transaction_handle *, | |
2518 | struct super_block *); | |
2519 | int journal_end(struct reiserfs_transaction_handle *, struct super_block *, | |
2520 | unsigned long); | |
2521 | int journal_end_sync(struct reiserfs_transaction_handle *, struct super_block *, | |
2522 | unsigned long); | |
2523 | int journal_mark_freed(struct reiserfs_transaction_handle *, | |
2524 | struct super_block *, b_blocknr_t blocknr); | |
2525 | int journal_transaction_should_end(struct reiserfs_transaction_handle *, int); | |
2526 | int reiserfs_in_journal(struct super_block *sb, unsigned int bmap_nr, | |
2527 | int bit_nr, int searchall, b_blocknr_t *next); | |
2528 | int journal_begin(struct reiserfs_transaction_handle *, | |
2529 | struct super_block *sb, unsigned long); | |
2530 | int journal_join_abort(struct reiserfs_transaction_handle *, | |
2531 | struct super_block *sb, unsigned long); | |
2532 | void reiserfs_abort_journal(struct super_block *sb, int errno); | |
2533 | void reiserfs_abort(struct super_block *sb, int errno, const char *fmt, ...); | |
2534 | int reiserfs_allocate_list_bitmaps(struct super_block *s, | |
2535 | struct reiserfs_list_bitmap *, unsigned int); | |
2536 | ||
033369d1 | 2537 | void reiserfs_schedule_old_flush(struct super_block *s); |
f466c6fd AV |
2538 | void add_save_link(struct reiserfs_transaction_handle *th, |
2539 | struct inode *inode, int truncate); | |
2540 | int remove_save_link(struct inode *inode, int truncate); | |
2541 | ||
2542 | /* objectid.c */ | |
2543 | __u32 reiserfs_get_unused_objectid(struct reiserfs_transaction_handle *th); | |
2544 | void reiserfs_release_objectid(struct reiserfs_transaction_handle *th, | |
2545 | __u32 objectid_to_release); | |
2546 | int reiserfs_convert_objectid_map_v1(struct super_block *); | |
2547 | ||
2548 | /* stree.c */ | |
2549 | int B_IS_IN_TREE(const struct buffer_head *); | |
2550 | extern void copy_item_head(struct item_head *to, | |
2551 | const struct item_head *from); | |
2552 | ||
2553 | // first key is in cpu form, second - le | |
2554 | extern int comp_short_keys(const struct reiserfs_key *le_key, | |
2555 | const struct cpu_key *cpu_key); | |
2556 | extern void le_key2cpu_key(struct cpu_key *to, const struct reiserfs_key *from); | |
2557 | ||
2558 | // both are in le form | |
2559 | extern int comp_le_keys(const struct reiserfs_key *, | |
2560 | const struct reiserfs_key *); | |
2561 | extern int comp_short_le_keys(const struct reiserfs_key *, | |
2562 | const struct reiserfs_key *); | |
2563 | ||
2564 | // | |
2565 | // get key version from on disk key - kludge | |
2566 | // | |
2567 | static inline int le_key_version(const struct reiserfs_key *key) | |
2568 | { | |
2569 | int type; | |
2570 | ||
2571 | type = offset_v2_k_type(&(key->u.k_offset_v2)); | |
2572 | if (type != TYPE_DIRECT && type != TYPE_INDIRECT | |
2573 | && type != TYPE_DIRENTRY) | |
2574 | return KEY_FORMAT_3_5; | |
2575 | ||
2576 | return KEY_FORMAT_3_6; | |
2577 | ||
2578 | } | |
2579 | ||
2580 | static inline void copy_key(struct reiserfs_key *to, | |
2581 | const struct reiserfs_key *from) | |
2582 | { | |
2583 | memcpy(to, from, KEY_SIZE); | |
2584 | } | |
2585 | ||
2586 | int comp_items(const struct item_head *stored_ih, const struct treepath *path); | |
2587 | const struct reiserfs_key *get_rkey(const struct treepath *chk_path, | |
2588 | const struct super_block *sb); | |
2589 | int search_by_key(struct super_block *, const struct cpu_key *, | |
2590 | struct treepath *, int); | |
2591 | #define search_item(s,key,path) search_by_key (s, key, path, DISK_LEAF_NODE_LEVEL) | |
2592 | int search_for_position_by_key(struct super_block *sb, | |
2593 | const struct cpu_key *cpu_key, | |
2594 | struct treepath *search_path); | |
2595 | extern void decrement_bcount(struct buffer_head *bh); | |
2596 | void decrement_counters_in_path(struct treepath *search_path); | |
2597 | void pathrelse(struct treepath *search_path); | |
2598 | int reiserfs_check_path(struct treepath *p); | |
2599 | void pathrelse_and_restore(struct super_block *s, struct treepath *search_path); | |
2600 | ||
2601 | int reiserfs_insert_item(struct reiserfs_transaction_handle *th, | |
2602 | struct treepath *path, | |
2603 | const struct cpu_key *key, | |
2604 | struct item_head *ih, | |
2605 | struct inode *inode, const char *body); | |
2606 | ||
2607 | int reiserfs_paste_into_item(struct reiserfs_transaction_handle *th, | |
2608 | struct treepath *path, | |
2609 | const struct cpu_key *key, | |
2610 | struct inode *inode, | |
2611 | const char *body, int paste_size); | |
2612 | ||
2613 | int reiserfs_cut_from_item(struct reiserfs_transaction_handle *th, | |
2614 | struct treepath *path, | |
2615 | struct cpu_key *key, | |
2616 | struct inode *inode, | |
2617 | struct page *page, loff_t new_file_size); | |
2618 | ||
2619 | int reiserfs_delete_item(struct reiserfs_transaction_handle *th, | |
2620 | struct treepath *path, | |
2621 | const struct cpu_key *key, | |
2622 | struct inode *inode, struct buffer_head *un_bh); | |
2623 | ||
2624 | void reiserfs_delete_solid_item(struct reiserfs_transaction_handle *th, | |
2625 | struct inode *inode, struct reiserfs_key *key); | |
2626 | int reiserfs_delete_object(struct reiserfs_transaction_handle *th, | |
2627 | struct inode *inode); | |
2628 | int reiserfs_do_truncate(struct reiserfs_transaction_handle *th, | |
2629 | struct inode *inode, struct page *, | |
2630 | int update_timestamps); | |
2631 | ||
2632 | #define i_block_size(inode) ((inode)->i_sb->s_blocksize) | |
2633 | #define file_size(inode) ((inode)->i_size) | |
2634 | #define tail_size(inode) (file_size (inode) & (i_block_size (inode) - 1)) | |
2635 | ||
2636 | #define tail_has_to_be_packed(inode) (have_large_tails ((inode)->i_sb)?\ | |
2637 | !STORE_TAIL_IN_UNFM_S1(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):have_small_tails ((inode)->i_sb)?!STORE_TAIL_IN_UNFM_S2(file_size (inode), tail_size(inode), inode->i_sb->s_blocksize):0 ) | |
2638 | ||
2639 | void padd_item(char *item, int total_length, int length); | |
2640 | ||
2641 | /* inode.c */ | |
2642 | /* args for the create parameter of reiserfs_get_block */ | |
2643 | #define GET_BLOCK_NO_CREATE 0 /* don't create new blocks or convert tails */ | |
2644 | #define GET_BLOCK_CREATE 1 /* add anything you need to find block */ | |
2645 | #define GET_BLOCK_NO_HOLE 2 /* return -ENOENT for file holes */ | |
2646 | #define GET_BLOCK_READ_DIRECT 4 /* read the tail if indirect item not found */ | |
2647 | #define GET_BLOCK_NO_IMUX 8 /* i_mutex is not held, don't preallocate */ | |
2648 | #define GET_BLOCK_NO_DANGLE 16 /* don't leave any transactions running */ | |
2649 | ||
2650 | void reiserfs_read_locked_inode(struct inode *inode, | |
2651 | struct reiserfs_iget_args *args); | |
2652 | int reiserfs_find_actor(struct inode *inode, void *p); | |
2653 | int reiserfs_init_locked_inode(struct inode *inode, void *p); | |
2654 | void reiserfs_evict_inode(struct inode *inode); | |
2655 | int reiserfs_write_inode(struct inode *inode, struct writeback_control *wbc); | |
2656 | int reiserfs_get_block(struct inode *inode, sector_t block, | |
2657 | struct buffer_head *bh_result, int create); | |
2658 | struct dentry *reiserfs_fh_to_dentry(struct super_block *sb, struct fid *fid, | |
2659 | int fh_len, int fh_type); | |
2660 | struct dentry *reiserfs_fh_to_parent(struct super_block *sb, struct fid *fid, | |
2661 | int fh_len, int fh_type); | |
b0b0382b AV |
2662 | int reiserfs_encode_fh(struct inode *inode, __u32 * data, int *lenp, |
2663 | struct inode *parent); | |
f466c6fd AV |
2664 | |
2665 | int reiserfs_truncate_file(struct inode *, int update_timestamps); | |
2666 | void make_cpu_key(struct cpu_key *cpu_key, struct inode *inode, loff_t offset, | |
2667 | int type, int key_length); | |
2668 | void make_le_item_head(struct item_head *ih, const struct cpu_key *key, | |
2669 | int version, | |
2670 | loff_t offset, int type, int length, int entry_count); | |
2671 | struct inode *reiserfs_iget(struct super_block *s, const struct cpu_key *key); | |
2672 | ||
2673 | struct reiserfs_security_handle; | |
2674 | int reiserfs_new_inode(struct reiserfs_transaction_handle *th, | |
2675 | struct inode *dir, umode_t mode, | |
2676 | const char *symname, loff_t i_size, | |
2677 | struct dentry *dentry, struct inode *inode, | |
2678 | struct reiserfs_security_handle *security); | |
2679 | ||
2680 | void reiserfs_update_sd_size(struct reiserfs_transaction_handle *th, | |
2681 | struct inode *inode, loff_t size); | |
2682 | ||
2683 | static inline void reiserfs_update_sd(struct reiserfs_transaction_handle *th, | |
2684 | struct inode *inode) | |
2685 | { | |
2686 | reiserfs_update_sd_size(th, inode, inode->i_size); | |
2687 | } | |
2688 | ||
2689 | void sd_attrs_to_i_attrs(__u16 sd_attrs, struct inode *inode); | |
2690 | void i_attrs_to_sd_attrs(struct inode *inode, __u16 * sd_attrs); | |
2691 | int reiserfs_setattr(struct dentry *dentry, struct iattr *attr); | |
2692 | ||
2693 | int __reiserfs_write_begin(struct page *page, unsigned from, unsigned len); | |
2694 | ||
2695 | /* namei.c */ | |
2696 | void set_de_name_and_namelen(struct reiserfs_dir_entry *de); | |
2697 | int search_by_entry_key(struct super_block *sb, const struct cpu_key *key, | |
2698 | struct treepath *path, struct reiserfs_dir_entry *de); | |
2699 | struct dentry *reiserfs_get_parent(struct dentry *); | |
2700 | ||
2701 | #ifdef CONFIG_REISERFS_PROC_INFO | |
2702 | int reiserfs_proc_info_init(struct super_block *sb); | |
2703 | int reiserfs_proc_info_done(struct super_block *sb); | |
2704 | int reiserfs_proc_info_global_init(void); | |
2705 | int reiserfs_proc_info_global_done(void); | |
2706 | ||
2707 | #define PROC_EXP( e ) e | |
2708 | ||
2709 | #define __PINFO( sb ) REISERFS_SB(sb) -> s_proc_info_data | |
2710 | #define PROC_INFO_MAX( sb, field, value ) \ | |
2711 | __PINFO( sb ).field = \ | |
2712 | max( REISERFS_SB( sb ) -> s_proc_info_data.field, value ) | |
2713 | #define PROC_INFO_INC( sb, field ) ( ++ ( __PINFO( sb ).field ) ) | |
2714 | #define PROC_INFO_ADD( sb, field, val ) ( __PINFO( sb ).field += ( val ) ) | |
2715 | #define PROC_INFO_BH_STAT( sb, bh, level ) \ | |
2716 | PROC_INFO_INC( sb, sbk_read_at[ ( level ) ] ); \ | |
2717 | PROC_INFO_ADD( sb, free_at[ ( level ) ], B_FREE_SPACE( bh ) ); \ | |
2718 | PROC_INFO_ADD( sb, items_at[ ( level ) ], B_NR_ITEMS( bh ) ) | |
2719 | #else | |
2720 | static inline int reiserfs_proc_info_init(struct super_block *sb) | |
2721 | { | |
2722 | return 0; | |
2723 | } | |
2724 | ||
2725 | static inline int reiserfs_proc_info_done(struct super_block *sb) | |
2726 | { | |
2727 | return 0; | |
2728 | } | |
2729 | ||
2730 | static inline int reiserfs_proc_info_global_init(void) | |
2731 | { | |
2732 | return 0; | |
2733 | } | |
2734 | ||
2735 | static inline int reiserfs_proc_info_global_done(void) | |
2736 | { | |
2737 | return 0; | |
2738 | } | |
2739 | ||
2740 | #define PROC_EXP( e ) | |
2741 | #define VOID_V ( ( void ) 0 ) | |
2742 | #define PROC_INFO_MAX( sb, field, value ) VOID_V | |
2743 | #define PROC_INFO_INC( sb, field ) VOID_V | |
2744 | #define PROC_INFO_ADD( sb, field, val ) VOID_V | |
2745 | #define PROC_INFO_BH_STAT(sb, bh, n_node_level) VOID_V | |
2746 | #endif | |
2747 | ||
2748 | /* dir.c */ | |
2749 | extern const struct inode_operations reiserfs_dir_inode_operations; | |
2750 | extern const struct inode_operations reiserfs_symlink_inode_operations; | |
2751 | extern const struct inode_operations reiserfs_special_inode_operations; | |
2752 | extern const struct file_operations reiserfs_dir_operations; | |
cd62cdae | 2753 | int reiserfs_readdir_inode(struct inode *, struct dir_context *); |
f466c6fd AV |
2754 | |
2755 | /* tail_conversion.c */ | |
2756 | int direct2indirect(struct reiserfs_transaction_handle *, struct inode *, | |
2757 | struct treepath *, struct buffer_head *, loff_t); | |
2758 | int indirect2direct(struct reiserfs_transaction_handle *, struct inode *, | |
2759 | struct page *, struct treepath *, const struct cpu_key *, | |
2760 | loff_t, char *); | |
2761 | void reiserfs_unmap_buffer(struct buffer_head *); | |
2762 | ||
2763 | /* file.c */ | |
2764 | extern const struct inode_operations reiserfs_file_inode_operations; | |
2765 | extern const struct file_operations reiserfs_file_operations; | |
2766 | extern const struct address_space_operations reiserfs_address_space_operations; | |
2767 | ||
2768 | /* fix_nodes.c */ | |
2769 | ||
2770 | int fix_nodes(int n_op_mode, struct tree_balance *tb, | |
2771 | struct item_head *ins_ih, const void *); | |
2772 | void unfix_nodes(struct tree_balance *); | |
2773 | ||
2774 | /* prints.c */ | |
2775 | void __reiserfs_panic(struct super_block *s, const char *id, | |
2776 | const char *function, const char *fmt, ...) | |
2777 | __attribute__ ((noreturn)); | |
2778 | #define reiserfs_panic(s, id, fmt, args...) \ | |
2779 | __reiserfs_panic(s, id, __func__, fmt, ##args) | |
2780 | void __reiserfs_error(struct super_block *s, const char *id, | |
2781 | const char *function, const char *fmt, ...); | |
2782 | #define reiserfs_error(s, id, fmt, args...) \ | |
2783 | __reiserfs_error(s, id, __func__, fmt, ##args) | |
2784 | void reiserfs_info(struct super_block *s, const char *fmt, ...); | |
2785 | void reiserfs_debug(struct super_block *s, int level, const char *fmt, ...); | |
2786 | void print_indirect_item(struct buffer_head *bh, int item_num); | |
2787 | void store_print_tb(struct tree_balance *tb); | |
2788 | void print_cur_tb(char *mes); | |
2789 | void print_de(struct reiserfs_dir_entry *de); | |
2790 | void print_bi(struct buffer_info *bi, char *mes); | |
2791 | #define PRINT_LEAF_ITEMS 1 /* print all items */ | |
2792 | #define PRINT_DIRECTORY_ITEMS 2 /* print directory items */ | |
2793 | #define PRINT_DIRECT_ITEMS 4 /* print contents of direct items */ | |
2794 | void print_block(struct buffer_head *bh, ...); | |
2795 | void print_bmap(struct super_block *s, int silent); | |
2796 | void print_bmap_block(int i, char *data, int size, int silent); | |
2797 | /*void print_super_block (struct super_block * s, char * mes);*/ | |
2798 | void print_objectid_map(struct super_block *s); | |
2799 | void print_block_head(struct buffer_head *bh, char *mes); | |
2800 | void check_leaf(struct buffer_head *bh); | |
2801 | void check_internal(struct buffer_head *bh); | |
2802 | void print_statistics(struct super_block *s); | |
2803 | char *reiserfs_hashname(int code); | |
2804 | ||
2805 | /* lbalance.c */ | |
2806 | int leaf_move_items(int shift_mode, struct tree_balance *tb, int mov_num, | |
2807 | int mov_bytes, struct buffer_head *Snew); | |
2808 | int leaf_shift_left(struct tree_balance *tb, int shift_num, int shift_bytes); | |
2809 | int leaf_shift_right(struct tree_balance *tb, int shift_num, int shift_bytes); | |
2810 | void leaf_delete_items(struct buffer_info *cur_bi, int last_first, int first, | |
2811 | int del_num, int del_bytes); | |
2812 | void leaf_insert_into_buf(struct buffer_info *bi, int before, | |
2813 | struct item_head *inserted_item_ih, | |
2814 | const char *inserted_item_body, int zeros_number); | |
2815 | void leaf_paste_in_buffer(struct buffer_info *bi, int pasted_item_num, | |
2816 | int pos_in_item, int paste_size, const char *body, | |
2817 | int zeros_number); | |
2818 | void leaf_cut_from_buffer(struct buffer_info *bi, int cut_item_num, | |
2819 | int pos_in_item, int cut_size); | |
2820 | void leaf_paste_entries(struct buffer_info *bi, int item_num, int before, | |
2821 | int new_entry_count, struct reiserfs_de_head *new_dehs, | |
2822 | const char *records, int paste_size); | |
2823 | /* ibalance.c */ | |
2824 | int balance_internal(struct tree_balance *, int, int, struct item_head *, | |
2825 | struct buffer_head **); | |
2826 | ||
2827 | /* do_balance.c */ | |
2828 | void do_balance_mark_leaf_dirty(struct tree_balance *tb, | |
2829 | struct buffer_head *bh, int flag); | |
2830 | #define do_balance_mark_internal_dirty do_balance_mark_leaf_dirty | |
2831 | #define do_balance_mark_sb_dirty do_balance_mark_leaf_dirty | |
2832 | ||
2833 | void do_balance(struct tree_balance *tb, struct item_head *ih, | |
2834 | const char *body, int flag); | |
2835 | void reiserfs_invalidate_buffer(struct tree_balance *tb, | |
2836 | struct buffer_head *bh); | |
2837 | ||
2838 | int get_left_neighbor_position(struct tree_balance *tb, int h); | |
2839 | int get_right_neighbor_position(struct tree_balance *tb, int h); | |
2840 | void replace_key(struct tree_balance *tb, struct buffer_head *, int, | |
2841 | struct buffer_head *, int); | |
2842 | void make_empty_node(struct buffer_info *); | |
2843 | struct buffer_head *get_FEB(struct tree_balance *); | |
2844 | ||
2845 | /* bitmap.c */ | |
2846 | ||
2847 | /* structure contains hints for block allocator, and it is a container for | |
2848 | * arguments, such as node, search path, transaction_handle, etc. */ | |
2849 | struct __reiserfs_blocknr_hint { | |
2850 | struct inode *inode; /* inode passed to allocator, if we allocate unf. nodes */ | |
2851 | sector_t block; /* file offset, in blocks */ | |
2852 | struct in_core_key key; | |
2853 | struct treepath *path; /* search path, used by allocator to deternine search_start by | |
2854 | * various ways */ | |
2855 | struct reiserfs_transaction_handle *th; /* transaction handle is needed to log super blocks and | |
2856 | * bitmap blocks changes */ | |
2857 | b_blocknr_t beg, end; | |
2858 | b_blocknr_t search_start; /* a field used to transfer search start value (block number) | |
2859 | * between different block allocator procedures | |
2860 | * (determine_search_start() and others) */ | |
2861 | int prealloc_size; /* is set in determine_prealloc_size() function, used by underlayed | |
2862 | * function that do actual allocation */ | |
2863 | ||
2864 | unsigned formatted_node:1; /* the allocator uses different polices for getting disk space for | |
2865 | * formatted/unformatted blocks with/without preallocation */ | |
2866 | unsigned preallocate:1; | |
2867 | }; | |
2868 | ||
2869 | typedef struct __reiserfs_blocknr_hint reiserfs_blocknr_hint_t; | |
2870 | ||
2871 | int reiserfs_parse_alloc_options(struct super_block *, char *); | |
2872 | void reiserfs_init_alloc_options(struct super_block *s); | |
2873 | ||
2874 | /* | |
2875 | * given a directory, this will tell you what packing locality | |
2876 | * to use for a new object underneat it. The locality is returned | |
2877 | * in disk byte order (le). | |
2878 | */ | |
2879 | __le32 reiserfs_choose_packing(struct inode *dir); | |
2880 | ||
ea0856cd | 2881 | void show_alloc_options(struct seq_file *seq, struct super_block *s); |
f466c6fd AV |
2882 | int reiserfs_init_bitmap_cache(struct super_block *sb); |
2883 | void reiserfs_free_bitmap_cache(struct super_block *sb); | |
2884 | void reiserfs_cache_bitmap_metadata(struct super_block *sb, struct buffer_head *bh, struct reiserfs_bitmap_info *info); | |
2885 | struct buffer_head *reiserfs_read_bitmap_block(struct super_block *sb, unsigned int bitmap); | |
2886 | int is_reusable(struct super_block *s, b_blocknr_t block, int bit_value); | |
2887 | void reiserfs_free_block(struct reiserfs_transaction_handle *th, struct inode *, | |
2888 | b_blocknr_t, int for_unformatted); | |
2889 | int reiserfs_allocate_blocknrs(reiserfs_blocknr_hint_t *, b_blocknr_t *, int, | |
2890 | int); | |
2891 | static inline int reiserfs_new_form_blocknrs(struct tree_balance *tb, | |
2892 | b_blocknr_t * new_blocknrs, | |
2893 | int amount_needed) | |
2894 | { | |
2895 | reiserfs_blocknr_hint_t hint = { | |
2896 | .th = tb->transaction_handle, | |
2897 | .path = tb->tb_path, | |
2898 | .inode = NULL, | |
2899 | .key = tb->key, | |
2900 | .block = 0, | |
2901 | .formatted_node = 1 | |
2902 | }; | |
2903 | return reiserfs_allocate_blocknrs(&hint, new_blocknrs, amount_needed, | |
2904 | 0); | |
2905 | } | |
2906 | ||
2907 | static inline int reiserfs_new_unf_blocknrs(struct reiserfs_transaction_handle | |
2908 | *th, struct inode *inode, | |
2909 | b_blocknr_t * new_blocknrs, | |
2910 | struct treepath *path, | |
2911 | sector_t block) | |
2912 | { | |
2913 | reiserfs_blocknr_hint_t hint = { | |
2914 | .th = th, | |
2915 | .path = path, | |
2916 | .inode = inode, | |
2917 | .block = block, | |
2918 | .formatted_node = 0, | |
2919 | .preallocate = 0 | |
2920 | }; | |
2921 | return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0); | |
2922 | } | |
2923 | ||
2924 | #ifdef REISERFS_PREALLOCATE | |
2925 | static inline int reiserfs_new_unf_blocknrs2(struct reiserfs_transaction_handle | |
2926 | *th, struct inode *inode, | |
2927 | b_blocknr_t * new_blocknrs, | |
2928 | struct treepath *path, | |
2929 | sector_t block) | |
2930 | { | |
2931 | reiserfs_blocknr_hint_t hint = { | |
2932 | .th = th, | |
2933 | .path = path, | |
2934 | .inode = inode, | |
2935 | .block = block, | |
2936 | .formatted_node = 0, | |
2937 | .preallocate = 1 | |
2938 | }; | |
2939 | return reiserfs_allocate_blocknrs(&hint, new_blocknrs, 1, 0); | |
2940 | } | |
2941 | ||
2942 | void reiserfs_discard_prealloc(struct reiserfs_transaction_handle *th, | |
2943 | struct inode *inode); | |
2944 | void reiserfs_discard_all_prealloc(struct reiserfs_transaction_handle *th); | |
2945 | #endif | |
2946 | ||
2947 | /* hashes.c */ | |
2948 | __u32 keyed_hash(const signed char *msg, int len); | |
2949 | __u32 yura_hash(const signed char *msg, int len); | |
2950 | __u32 r5_hash(const signed char *msg, int len); | |
2951 | ||
2952 | #define reiserfs_set_le_bit __set_bit_le | |
2953 | #define reiserfs_test_and_set_le_bit __test_and_set_bit_le | |
2954 | #define reiserfs_clear_le_bit __clear_bit_le | |
2955 | #define reiserfs_test_and_clear_le_bit __test_and_clear_bit_le | |
2956 | #define reiserfs_test_le_bit test_bit_le | |
2957 | #define reiserfs_find_next_zero_le_bit find_next_zero_bit_le | |
2958 | ||
2959 | /* sometimes reiserfs_truncate may require to allocate few new blocks | |
2960 | to perform indirect2direct conversion. People probably used to | |
2961 | think, that truncate should work without problems on a filesystem | |
2962 | without free disk space. They may complain that they can not | |
2963 | truncate due to lack of free disk space. This spare space allows us | |
2964 | to not worry about it. 500 is probably too much, but it should be | |
2965 | absolutely safe */ | |
2966 | #define SPARE_SPACE 500 | |
2967 | ||
2968 | /* prototypes from ioctl.c */ | |
2969 | long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); | |
2970 | long reiserfs_compat_ioctl(struct file *filp, | |
2971 | unsigned int cmd, unsigned long arg); | |
2972 | int reiserfs_unpack(struct inode *inode, struct file *filp); |